The New Panel That Tripped Every AFCI

Why this matters

A panel changeout in an older house moves every branch circuit onto new breakers, and when those breakers are AFCIs most of them do something the old ones never did: they compare the current going out on the hot with the current coming back on that breaker's own neutral. Any wiring quirk the old panel tolerated for fifty years, a neutral shared between two circuits, a white landed one terminal over, becomes an instant trip. This composite case is a changeout where half the new AFCIs would not hold on the first energize, worked through by elimination, and the one step the shop added to its transfer procedure afterwards.

The symptom

A 1960s house, the old fuse-and-breaker panel replaced with a new 200 A main-breaker panel in the same location. Twenty-four branch circuits transferred; sixteen of them, on the jurisdiction's reading of NEC 210.12, NEC 2023 (2020 where the section moved), as adopted, went onto single-pole combination AFCI breakers because their conductors had been extended to reach the new positions. (210.12(D) excuses an extension of 6 ft or less that adds no outlets or devices; this AHJ treated the changeout as a modification needing AFCI, so check how yours reads it.)

At the end of the day the crew closed the main and then every branch breaker. Eight of the sixteen AFCIs tripped within seconds: positions 5, 7, 9, 11, 15, 17, 19 and 21. The other eight held, and so did every non-AFCI breaker. Several of the tripped breakers showed a ground-fault indication on their trip indicators rather than an arc indication, which the crew noted and did not yet trust.

The homeowner had a refrigerator on one of the tripped circuits and a family coming home at six. The lead moved the refrigerator to a temporary cord from a holding circuit and started working the problem with the clock visible.

First suspect: a bad batch of breakers

Eight failures out of sixteen from the same carton is the kind of number that makes a crew blame the product. The test is cheap: move a tripping circuit onto a breaker that has just held.

Main off, bus proved dead live-dead-live with a tester proven on a known live source before and after, per NFPA 70E Article 120 (120.5 in the 2021 edition, renumbered in 2024), work practices under 29 CFR 1910.333(b)(2). The line-side lugs stay energized from the utility with the main open, so they were covered with the panel's barrier before anyone reached in. Whatever was landed on breaker 7 was swapped with what was landed on breaker 13, whose AFCI had held.

Re-energized from the side of the panel, deadfront on, and the two closed alone. Circuit 7, now on the "good" breaker, tripped. Circuit 13, now on the "bad" one, held. The trip followed the wiring, not the breaker. Moved back.

Second suspect: damaged old wiring

Fifty-year-old cable in a house that has been remodeled twice can have a nicked conductor or a neutral touching a box. A neutral-to-ground fault would also explain the ground-fault indications.

With the main off and the bus proved dead again, the eight neutrals were lifted from their breaker terminals, plug-in loads on those circuits were unplugged so surge strips with protective components between neutral and ground would not drag the reading, hardwired electronics on those circuits (smoke alarms, electronic dimmers, LED drivers) were disconnected so 500 V could not damage them, and each neutral was insulation-tested to the equipment ground at 500 V. All eight read over 200 megohms. See related: How to Insulation Test a Feeder or Branch Circuit.

No neutral touched ground. The ground-fault indication was real, but it was not coming from a fault in the cable.

Third suspect: the loads

Old fluorescent fixtures, a furnace blower and a bathroom fan can upset an AFCI's arc detection, and some crews start pulling plugs here. But on breakers that report a trip cause, a load-induced trip shows the arc indication, takes time to develop, and does not depend on which other breakers are on. These tripped within seconds with ground-fault indications. Rather than unplug the house, the lead tested the breakers in a way that would show whether they depended on each other.

One breaker at a time, then two

All AFCIs off. Each tripping circuit was closed alone with its lights on and one plug load running, then opened. Then pairs.

 AFCI TRIP LOG   panel changeout, 16 AFCI positions
 all closed together: 8 tripped
 5  7  9  11  15  17  19  21

 ALONE  (lights on + one plug load each)
 pos 5   held     pos 7   TRIP     pos 9   held
 pos 11  TRIP     pos 15  held     pos 17  TRIP
 pos 19  TRIP     pos 21  TRIP

 PAIRS  (first breaker on, then second closed)
 5 on, close 7     -> 5 and 7 both trip
 9 on, close 11    -> 9 and 11 both trip
 15 on, close 17   -> 15 and 17 both trip
 19 and 21         -> each trips alone, together too

 DEAD  N to EGC, 500 V, loads unplugged
 all 8 neutrals    -> over 200 megohm

The log sorted the eight into two different patterns, and that sort is the diagnosis.

Three pairs that held alone on one side and failed together. 5, 9 and 15 each held alone. 7, 11 and 17 each tripped alone. Closing the partner tripped both. A breaker that holds by itself but trips the moment one specific other breaker closes is sharing something with it, and the only thing two branch circuits share is a neutral.

One pair where each tripped alone. 19 and 21 tripped by themselves with nothing else on. Neither one's return current was coming back through its own breaker.

What a shared neutral does to two single-pole AFCIs

A multiwire branch circuit runs two hots on opposite legs with one neutral, usually in a 3-wire cable: black, red, white. The neutral carries the difference between the two hots' currents. On the old panel it landed on the neutral bar and nothing cared.

On the new panel each hot went to its own single-pole AFCI and the one white, as tagged, was landed on the load-neutral terminal of whichever of the two breakers the crew reached first: positions 5, 9 and 15. Each of those breakers then had its own hot out and a neutral that also carried the partner's return. Alone, it saw its own current out and its own current back, balanced, so it held. Its partner (7, 11, 17) had a hot and an empty neutral terminal: current out, nothing back through its sensor, so it read the whole load as leakage to ground and tripped, which is exactly what a ground-fault indication means. With both closed, each saw a mismatch equal to the other circuit's current, and both tripped.

The cables confirmed it with everything dead. In each of the three cases, a single 3-wire cable came into the panel with its black on one position and its red on the adjacent one. On the old panel those had been two single-pole breakers with no handle tie, which is why nobody had marked them as multiwire circuits. The tag on the white had been written for one breaker number, and the crew had faithfully landed it there.

The mislanded neutral

Positions 19 and 21 were two ordinary 2-wire circuits in separate cables. Tracing each white back along its own cable sheath into the panel showed that circuit 19's white was landed on breaker 21's load-neutral terminal. Circuit 21's white, finding its terminal taken, had been landed straight on the neutral bar.

One mislanded white explains both trips. Breaker 19 sends current out and gets it back through breaker 21. Breaker 21 sends current out and gets it back through the bar, bypassing its own sensor. Each sees an unbalanced current with nothing else on, and each trips alone.

The repair and the proof

Main off, bus proved dead, line-side barrier in place.

• The three multiwire circuits: each pair of single-pole AFCIs came out and a 2-pole AFCI listed for multiwire branch circuits went in, with the shared white on its one neutral terminal. That also satisfies NEC 210.4(B), which requires a means to disconnect both hots of a multiwire circuit simultaneously, and the conductors of each circuit were grouped with a cable tie in the panel per 210.4(D). Adjacent positions on a 2-pole breaker are on opposite legs by construction, so the shared neutral carries the difference of the two currents, not the sum.
• Circuits 19 and 21: 19's white moved to breaker 19's terminal, 21's white off the bar and onto breaker 21.

Every termination was torqued to the value marked on the breaker with a calibrated tool (110.14(D)), and the directory now names the three 2-pole breakers as multiwire circuits.

Energizing: deadfront on, the lead to the side of the panel, the homeowner and crew clear. Main closed, then each repaired circuit alone with its loads on, then everything together. All sixteen AFCI positions held, now thirteen breakers counting the three 2-pole units. Each AFCI then tripped on its own test button and reset. Clamped at the panel with the multiwire loads running, each 2-pole breaker's shared neutral read less than the larger of its two hot currents, which is what opposite legs should give. The refrigerator went back on its own circuit before six.

What the shop changed

The circuits were never the problem; the house had been wired that way for decades. The problem was that the transfer procedure tagged conductors one per breaker, which hides the fact that one white serves two.

The shop added a tag-before-cut step to How to Move Branch Circuits Into a Replacement Panel. Before any cable is cut free of the old panel, every conductor gets the old breaker number, and every cable carrying two ungrounded conductors and one white is tagged MWBC with both numbers on the white, unless it is proven to be a 240 V load or a switch loop. Every MWBC is ordered as a 2-pole AFCI before the job, not discovered at the end of it. On the next three changeouts that step found two more multiwire circuits, both before the crew drove to the supply house.

References

• NEC 2023 (2020 where the section moved), as adopted: 210.4(B), 210.4(D), 210.12, 110.14(D)
• NFPA 70E, Article 120 (120.5 in the 2021 edition, renumbered in 2024), current edition as your safety program adopts it
• 29 CFR 1910.333(b)(2), deenergized parts
• See related: How to Move Branch Circuits Into a Replacement Panel
• See related: GFCI, AFCI, and DFCI Protection Selection by NEC Article