080806-1554 EST
mikekelly:
First, the subpanel is wired wrong as has been stated.
Second, to answer your questions, and a couple more.
Let's assume we have only a 0-120 subpanel, no 240.
How does a GFCI work?
Internally there is no connection of the EGC terminal to any of the GFCI components. Thus, other than for an external tester the EGC terminal and EGC sockets have nothing to do with the GFCI operation.
Internal to the GFCI the hot and neutral conductors pass thru a current transformer. If there is no external shunt path or current source on the load side of the current transformer, then the current transformer output is zero because the neutral and hot conductor currents are exactly equal but of opposite direction and their magnetic fields cancel.
If you unbalance these currents by approximately 5 MA, then a threshold detector will trip the GFCI device. There is filtering on the threshold operation so larger currents cause quicker tripping.
An aside. If I apply a DC unbalance of a reasonable value the GFCI will not trip. In other words considerably above the nominal 5 MA threshold. A dangerous condition that would not cause a trip is: charging a 120 V battery with the negative terminal of the battery connected to neutral, and a short from +120 V to EGC thru your body.
A basic understanding of how GFCI devices work and electrical theory provides a means to answer your questions. I suggest you take one apart and study its construction and operation.
I believe your circuit has neutral and EGC connected together at the subpanel, but isolated from the panel. Assume nothing connects to the subpanel enclosure. Also nowhere beyond the subpanel is there any connection between the EGC and neutral.
What would happen if there was a ground fault?
An adequate conductive path from hot or neutral on the load side of the GFCI to EGC will trip the GFCI. If the short was from neutral to EGC, then there needs to be some load to cause unbalanced current flow.
Anything that will cause an unbalanced current thru the current transformer that exceeds the threshold will trip the GFCI. If I have a 500 V AC supply somewhere and referenced to earth anywhere or in the case of neutral before the GFCI and an appropriate resistor from the 500 V to either the load side neutral or hot wire will trip the GFCI.l
What would happen if there was a an open neutral?
An open neutral to a GFCI receptacle will deprive the GFCI of power to trip the GFCI. Thus, a short of a GFCI output hot line to EGC will not trip the GFCI. Nothing else will trip the device except the TEST button (Leviton device I analyzed). Loss of neutral after the GFCI (on the load side) will deprive the load of power. Nothing will happen until something causes an unbalanced current thru the GFCI, and then it will trip.
Does the GFCI need a equipment ground to properly function?
No. If by that you mean will an unbalanced current thru the current transformer cause the GFCI to trip. This of course requires input power to neutral and hot of the GFCI.
The Leviton can not be RESET (circuit closed and latched) if there is no neutral-hot input power. The TEST button only mechanically trips the latch and I do not call that a test. However, on pushing the RESET button a test is performed and if this is not successful, then the GFCI contacts do not close, and you can not latch the GFCI. You must have input power to RESET and latch the GFCI. Connecting the input neutral and hot to the ouput terminals will not allow you to RESET and latch the GFCI.
If you latch the GFCI, requires correctly connected power and pushing the RESET button, then remove power the latch will remain latched. Now disconnect the power source, switch the power leads to the output terminals, and you will get power flow thru the GFCI. An unbalanced current should trip the GFCI. I have not tested this. Once tripped it would not be possible to RESET the GFCI.
My above assumption that the subpanel enclosure is floating, not connected to anything, means if you short the hot output of a GFCI receptacle to the enclosure that may not trip the GFCI. It will make the subpanel enclosure dangerous.
See my posts 162 and 171 an evolution in analysis at
http://forums.mikeholt.com/showthread.php?t=102022&page=5
.