I am repairing some underwater lights in a pond. They are line voltage and are from the Roman lighting company. They are sealed lights that come with a whip attached. They mount with a set screw on a post and shine up on a sign under a waterfall.

This is new ground for me. Is it required that they be gfci protected? According to 680.23(3) they need to be protected during relamping. They cannot be energized when they are out of the water and cannot be relamped when under water.

Is this a natural waterfall or pumped water?
If it is pumped water you have a fountain and 680.51 applies



If this is a natural pond and waterfall, it is not clear you need a GFCI for an underwater light, only a receptacle. See article 682

Greg,

I myself would try to install the lights on a GFI just to err on the side of safety. However that is just my 2 cents.

I agree, anything outside should be GFCI IMHO but there are still some holes in NFPA70.

I didi the repair today. The original install was three circuits, each to a double gang weather box at the pond. They had put a gfci dead front in the box with a blank cover over it, then caulked the dickens out of it. This was ok until the gfci tripped. I was an easy fix. I put in use covers on the boxes.

I tried gfci breakers. That seemed to be the best set-up, however they wouldn't hold.

Hmmmmm....

"I tried gfci breakers. That seemed to be the best set-up, however they wouldn't hold."

Wonder why?

I imagine the cable/raceway to the pond is damaged or a J box along the way is wet.

It is very common for long runs to trip GFCI devices even when there is no damage to the wiring. The capacitive effect of the conductors running adjacent to each other plus the inrush of the load is often more than enough to cause nuisance tripping. The closer the GFCI is to the load, the less nuisance tripping.

Another possibility is if the wiring shares a conduit or is closely spaced with other circuits (not in conduits) when leaving the panelboard (even for a short distance), the breaker can be more sensitive due to coupling between adjacent circuit wiring, especially if there is high frequency noise generated by computer and/or solid state ballast loads.

Any type of inductive load such as magnetic ballasts or small motors at a distance from the GFCI device will also cause tripping.

This type of nuisance tripping has largely been ignored by the manufacturers. The testing they do is primarily at their labs and they are not out in service trucks, so they don't see a problem.