Every so often, roughly a 1/20 chance, flipping the light switch to my laundry room's fluorescent fixture will trip the nearby GFCI outlet. Why is this? Is it more likely to be the GFCI being faulty or the fixture itself, like a faulty ballast? Or could it be the light switch? The switch has already been recently replaced, but I don't think I grounded it during installation (it wasn't grounded initially so I left it that way). Could having a light switch that is not grounded cause this phenomenon? The fixture is probably older than 20 years. It uses a double circline bulb. Thanks.
The problem could be caused by Electromagnetic interference, or an attempt to filter the interference.
Older Ground-fault circuit interrupting (GFCI) devices may be more susceptible to EMI related nuisance tripping. If you have an older GFCI device, you should first try replacing the GFCI device.
Some fluorescent fixtures have an EMI/RFI filter built into the device, to try and prevent any interference produced by the fixture (ballast or bulbs) from leaving the fixture. In most cases the interference is bled off through the equipment grounding conductor (EGC), which can lead to an imbalance of current between the ungrounded (hot) conductor and the grounded (neutral). This current imbalance may be large enough, and long enough to trip a GFCI device.
It's quite possible that there may actually be a ground-fault somewhere in the fixture, or the switch. Could be a stray arc somewhere in the fixture when it turns on/off, that causes a small amount of current to leak. It might be worth trying a more modern fixture.
Why is it GFCI protected?
You may want to figure out why the fixture is GFCI protected in the first place. If the protection is not required, removing the GFCI protection will certainly solve your problem.
The bit about the light shutting off when the GFCI trips is very important, as it tells us that the light is connected to the load side of the GFCI outlet.
What this means is the outlet is also protecting the light. This is a little overkill. It may be possible to connect it to the line side of the outlet with the other power wires. This should fix your current problem.
Usually outlets and lights are on different circuits, so that overloading an outlet won't shut off the lights. You may want to see about doing that in the future.
The reason the light trips the light in your current configuration is that CFL lights take a lot of energy to start up. This energy takes the form of a spike, which can cause the power going through the conductors to be imbalanced for a tiny fraction of a second. GFCI devices are designed to trip if they see an imbalance. The reason it doesn't always trip is that there is a delay between when the GFCI device first sees this imbalance, and when it trips. If it didn't, just about everything more complex than an incandescent light bulb would trip them.
I have been told by an engineer who designs these, that an inductive inrush current on the line conductor can and will trip the receptacle, it just depends upon the magnitude and time duration of the event. The newer production units are much faster than the older ones and monitor all 3 wired connections. This does apply to a transformer or in your case a lighting ballast. The iron core saturation time constant period does not allow a current to pass beyond it (remember that inductive filters react to a change in current and capacitive filters react to a change in voltage) until the iron core is fully saturated and thus fully mutually inducted - coupled to the secondary or output side. The GFCI sees this as an improper current flow and will operate. It assumes that an external return path for the current (IE a fault) has been found and is being used to complete the circuit and not the hard wired neutral path. Therefore an improper current flow of sufficent magnitude and duration on any of the 3 conductors will operate the device. My original question to him was why did a GFCI placed on the end of a 100ft long 10awg cord trip instantly upon plugging in the cord with or without a load connected to the far end. The answer was that the return (neutral) current for energizing the cord was of an adequate time duration that it was within the parameters of the device recognizing it as a fault current. Do the math, at lamba how long does it takes the initial sine wave to reach the GFCI at the end of the cord, Now the amount of time it takes to enable the logic, literally a null factor and last the same amount of time as the first equation to return the sine back to the neutral commoning point and completely balance the circuit. This last one is what messes up everything. As a check I made a 10ft cord with same GFCI at far end, it did not trip! This is why the large GFCI circuit breakers have adjustable trips. to fit the actual operational aspects of individual circuits.