GFCI and AFCI Nuisance Trip Service Visit

Purpose

Nuisance is a hypothesis, and this procedure stops the shop from acting on it. A protective device that trips is usually reporting something. The job is to measure what it is reporting, in milliamps or in megohms, before anything is replaced, because the two cheap wrong answers are both expensive: swapping the device when the fault is in a load, and blaming a load when the wiring is wet.

The specific failure this prevents is a tech who replaces a GFCI, hands over a working receptacle, and leaves a towel warmer with a wet element on the circuit. It trips again in a week, the shop returns, and the second visit starts from the same absence of measurement as the first. Two visits and a part later, nobody has a number.

Scope

Covers service calls where a ground-fault circuit interrupter, an arc-fault circuit interrupter or a dual-function device is opening in normal use, in dwellings and light commercial premises at 120 V and 240 V.

Does not cover a device that will not reset because a genuine short or ground fault is present downstream, which is a fault call rather than a tripping-device call. Does not cover the phone screen, which the intake SOP owns and which routes calls here.

Roles and responsibilities

Procedure

1. Set this up as a live-circuit measurement task and decide in advance what will not be done energized. Acceptance: the arc flash and shock assessment for the panel is complete before any breaker work, leakage measurement happens at the cord or the device with covers in place, and anything behind a deadfront follows the isolation SOP. Wrong looks like a deadfront off so a clamp will fit around a conductor. Stop rule: if a measurement needs an enclosure open, it becomes a de-energized task or an assessed energized one, never a quick look. Hazard: a leakage clamp is not an insulated tool and it puts a hand near a bus.

2. Establish the exact trip history before touching anything. Acceptance: you have how often, at what time of day, in what weather, with which appliance running, and whether it trips with everything unplugged. Wrong looks like "it just trips sometimes." Stop rule: if it trips only in rain or only after irrigation runs, treat it as a wet-location fault and go find water in a box rather than starting a load walk-down; calling that nuisance sends you to the wrong end of the circuit. Hazard: an outdoor box holding water is energized standing water, so the circuit is opened and locked before the cover comes off.

3. Confirm the instrument can resolve the quantity you are about to measure. Acceptance: the clamp meter has a dedicated leakage-current function with resolution of 0.01 mA or better and reads a known small current correctly. Wrong looks like a general-purpose clamp on a 60 A range showing 0.00 and a tech concluding there is no leakage. Stop rule: do not record a zero produced by an instrument that cannot resolve the quantity; stop and get the right meter. Hazard: none from the meter, but a false zero condemns the device, which is the wrong part.

4. Map the protected zone before assuming what it covers. Acceptance: a written list of every receptacle, fixture and hardwired load downstream, confirmed by testing each one, including outlets in other rooms. Wrong looks like assuming a bathroom GFCI protects only that bathroom, when the same device commonly feeds a garage, an exterior receptacle and a second bath. Stop rule: an unmapped zone makes every later measurement ambiguous, so the mapping finishes before any load is unplugged. Hazard: none, but leave every cover plate on; this is a plug-in tester walk, not a teardown.

5. Run the device's own test and its end-of-life check. Acceptance: pressing TEST opens the device, pressing RESET latches it, and it holds with the load connected. On a GFCI built to the self-test requirement added in the 2015 edition of UL 943, a device at end of life will refuse to reset or will indicate a fault. Wrong looks like a device that trips on TEST but will not latch. Stop rule: a device that will not latch is replaced first, with a listed replacement per NEC 406.4(D), before any further diagnosis, because a locked-out device cannot test anything downstream. Hazard: replacing a device is a de-energized task; the circuit is opened, locked and proved dead at the device.

6. Measure total circuit leakage with everything connected and the device reset. Acceptance: a clamp around the ungrounded and grounded conductors together, reading the imbalance, gives a stable value in milliamps that you write down. Wrong looks like clamping only the hot conductor and reading load current. Stop rule: a reading at or above the 4 mA lower edge of the UL 943 band means the device is operating correctly rather than failing, and the diagnosis moves to the loads; do not replace a device that is doing its job. Hazard: this is a live measurement, taken with the enclosure closed wherever the geometry allows.

7. Walk the loads down one at a time and attribute the leakage. Acceptance: after each disconnection the value is re-read and recorded, and the attributed values sum to the original total within the meter's resolution. Wrong looks like unplugging three things at once and getting a number nobody can attribute. Stop rule: if the sum does not close, something was missed at step 4 and you return to the mapping. Hazard: unplugging a load under current draw can arc at the plug face, so switch the appliance off at its own control first.

8. If the loads do not account for it, test the wiring for a crossed or contacting grounded conductor. Acceptance: with the device open, the protected circuit's grounded conductor is separated and confirmed to have no continuity to the panel neutral bus, to ground, or to another circuit's grounded conductor. Wrong looks like a white conductor with a low-resistance path to the enclosure. Stop rule: any such path is the fault; find it and fix it rather than reporting it as intermittent. Hazard: this step opens a grounded conductor, so both ungrounded conductors of any multiwire branch circuit are opened and locked first per NEC 210.4(B); opening a loaded neutral puts the return current through whoever bridges the gap.

9. Where the fault is still unattributed, run an insulation resistance test on the de-energized, load-disconnected circuit. Acceptance: with every electronic load, dimmer, surge device and the protective device itself disconnected, a 500 V DC test between conductors and to ground reads at or above a 1 megohm field floor. Treat that as a minimum, not a target: a healthy dry branch circuit commonly reads in the hundreds of megohms, so 2 megohms passes the floor and is still a finding worth chasing. Wrong looks like a low reading recorded with a television still plugged in. Stop rule: any single-digit megohm reading is reported and traced, not averaged away. Hazard: a megohmmeter puts 500 V DC on the conductors, destroying anything electronic left connected and leaving a capacitive charge on a long run, so use the discharge function and prove the conductors at zero before handling them.

10. Replace only what the measurement condemned, then re-test and leave a customer instruction. Acceptance: the device or load measured to be at fault is corrected, total leakage is re-measured and recorded, and the device holds through a deliberate exercise of the loads that used to trip it. Wrong looks like replacing the device and the receptacle and one appliance because that is easier than choosing. Stop rule: if the re-measured value has not moved, the fault is still on the circuit and the visit is not finished; say so on the ticket rather than closing it as resolved. Hazard: none, but tell the customer to test the device monthly by its own TEST button, since that is the only check that catches a device which has quietly stopped protecting.

The record this produces

One measurement record per protected zone. Fields: device type, make and manufacture date, the mapped zone with every outlet listed, the step 2 trip history, the meter used and its leakage resolution, total leakage with all loads, the attributed value for each load with the running total after each disconnection, any insulation resistance readings with test voltage and what was disconnected, the corrective action, and the re-measured total.

The attribution table has the longest life. When the same circuit calls again in a year, the previous walk-down tells the next tech which loads were already cleared and what the baseline residual was, which is often the difference between an hour and a morning. The supervising master reads the records where the sum did not close, because that is a mapping failure and mapping is a training issue rather than a diagnostic one. On AFCI calls, the record of which appliance was running at each trip is what separates a genuine series arc from a vacuum motor or an LED driver the device dislikes.

Worked pass: bathroom GFCI opening two or three times a week

Step 2 history: trips two or three times a week, more often in the morning, never overnight, and the customer suspects the hair dryer without being sure. Not weather-related, so the wet-location branch does not fire.

Step 3 fails. The tech's general-purpose clamp has no leakage function and reads 0.00 A on its lowest range with the circuit loaded. Under the stop rule that zero is not recorded, and the diagnostic pauses while the leakage clamp comes off the second truck. The temptation here is to write "no leakage found, replaced device," which is precisely the visit that produces a second visit.

Step 4 maps five outlets rather than the two the customer described: two bathroom receptacles, a cord-connected towel warmer behind the door, a lighted mirror, and an exterior receptacle on the far side of the wall.

Step 5 passes. The device trips on TEST, latches on RESET and holds.

Step 6 reads 4.2 mA total with everything connected. That sits inside the 4 to 6 mA band UL 943 sets, so the device is operating correctly and is not the part at fault.

Step 7 walks the loads down. Unplugging the towel warmer drops the reading to 1.6 mA, so it accounts for 2.6 mA. Unplugging the hair dryer, plugged in but switched off, drops it to 0.9 mA, so the dryer accounts for 0.7 mA. Unplugging the lighted mirror drops it to 0.4 mA, so the mirror accounts for 0.5 mA. The residual 0.4 mA is the wiring and the device. The attributed values sum to 2.6 plus 0.7 plus 0.5 plus 0.4, which is 4.2 mA and matches the total, so nothing on the circuit was missed.

The towel warmer alone is 2.6 of the 4.2 mA measured, roughly 62 percent of the total, and its element has been absorbing moisture. With it removed the circuit sits at 1.6 mA, comfortably under the 4 mA edge below which the device is not permitted to trip.

Steps 8 and 9 were not needed, and the record says so rather than leaving them blank. The customer was told the finding plainly: the receptacle is fine, the wiring is fine, and the appliance is the problem. Nobody enjoys that conversation, but it is the one the measurements support, and the alternative is a replaced device and a return visit.

References

• UL 943, ground-fault circuit interrupters, including the 4 to 6 mA trip band and the self-test requirement added in its 2015 edition
• UL 1699, arc-fault circuit interrupters; confirm any ground-fault trip level from the specific manufacturer's literature rather than assuming a common value
• NEC (NFPA 70) 210.8 and 210.12 for where these devices are required, 406.4(D) for replacement receptacles, and 210.4(B) for multiwire branch circuit disconnection
• NFPA 70E-2021, Article 130, for the assessment run at step 1 before any panel entry
• See related: the lockout tagout for branch circuit work SOP, which owns the isolation used at steps 8 and 9