Short answer: with the service neutral open, a ground fault at nominal leg voltage returns through the earth at about 2.4 A, and a 20 A breaker never moves. The bonded enclosure sits about 60 V above the soil while that fault burns; an unbonded one sits at the full line voltage. The single neutral-to-case bond at the service disconnect is what holds those numbers down. Walk the three wiring cases below, check your own panels against them, and fix stray bonds first.
What are the symptoms of a lost service neutral?
The signature is a voltage split that moves with the loads: a light-loaded leg climbs toward double its nominal voltage while a heavy-loaded leg collapses toward 0 V. Dimming lights and dying electronics on the hot leg follow within seconds.
The two hots turn into a voltage divider through the loads themselves. The midpoint drifts wherever the imbalance pushes it. Two legs loaded at 48 Ω and 480 Ω divide the line-to-line supply into about 22 V and 218 V. A power supply built for half the line voltage takes four times its rated heating when it sees the doubled value, so the hot leg fails fast. The split jumps every time a compressor starts.
- Bright lights on one leg and dim lights on the other, trading sides as loads cycle: the neutral point is floating.
- Electronics and LED drivers dying on one leg while the other leg recovers: the hot leg is running far over nominal.
- Water heaters, dryers and other line-to-line loads working normally: they never touch the neutral, so the garage looks fine while the kitchen complains.
- A steady line-to-line reading at every outlet while leg-to-neutral readings swing from near 0 to nearly the whole supply: the source is healthy, the return is not.
- Tingly enclosures, nuisance GFCI trips or warm conduit: neutral current is finding its way home through bonded metal.
Any of these right after a storm points at the utility drop first. The neutral splice at the pole or the weatherhead fails far more often than a conductor inside your walls.
Why won't the breaker clear a ground fault while the utility neutral is open?
Because the only path left to the transformer runs through the earth: with two 25 Ω electrodes in the loop, a fault at the nominal leg voltage draws about 2.4 A, and a 20 A breaker holds forever at 12% of its rating.
NEC 250.4(A)(5) says the same in code words: the effective ground-fault current path is the metal, and the earth is not considered part of it. Grounding electrodes hold the system near earth potential; they were never sized for fault current. With the utility neutral gone, the electrodes inherit the job, and their tens of ohms strangle the current. That is the whole argument of the grounding and bonding guide: fault current needs metal all the way home.
Worked example, with the assumptions on the table so you can substitute your own:
- Bonded 20 x 20 x 8 in stainless subpanel enclosure, 40 ft from the service disconnect.
- 12 AWG copper EGC at about 1.6 Ω per 1000 ft, so 0.06 Ω for the run.
- A 25 Ω electrode at each end, the familiar NEC figure for a single made rod, assumed at both the service and the transformer.
The loop is 25 Ω + 25 Ω + roughly 0.1 Ω of copper ≈ 50 Ω. Fault current: 120 V / 50 Ω = 2.4 A, which a 20 A breaker carries indefinitely. The case and everything bonded to it rise to 2.4 A x 25 Ω = 60 V above the soil. Bridge that with a conservative 1 kΩ wet-skin body and you carry 60 mA, four to six times the 10-16 mA let-go values from IEC 60479-1 and Dalziel's tests. Nothing trips, and the enclosure waits for a hand.
What does the faulted enclosure case sit at in each wiring case?
Three cases decide it: a bonded case rides at about 60 V during an open utility neutral, the same bond clears a downstream neutral break in one cycle at 240 A, and an unbonded case sits at the full leg voltage with no trip at all.
| Wiring case | Fault return | Fault current | Case voltage to earth | Breaker |
|---|---|---|---|---|
| Bond at the service disconnect, EGCs intact, utility neutral open | Copper to the service bar, then earth | about 2.4 A | about 60 V | holds |
| Same panel, feeder neutral open instead | Full copper loop, about 0.5 Ω | about 240 A | near 0 V | trips in about one cycle (17 ms) |
| No EGC and no bond | Nothing deliberate | leakage only | the full leg voltage | holds |
The middle row uses the same arithmetic: assume about 0.5 Ω of conductor round trip, a fair figure for a 100 ft feeder, and 120 V / 0.5 Ω = 240 A. That is twelve times a 20 A rating, which lands the breaker in its magnetic region and clears in about one cycle, 17 ms at 60 Hz. The bond earns nothing while the drop is healthy; it buys those 17 ms the day a neutral opens downstream of it. The transformer insulation failure case tells the same story from the faulted-equipment side.
Why is a neutral bond in a subpanel the worst case?
A neutral bond in a subpanel turns the enclosure into a current-carrying conductor: 10 A of unbalanced load across 0.05 Ω of box and raceway puts 0.5 V on every surface you can touch. With a neutral open, the case becomes the return for the whole panel.
NEC 250.6 aims to keep neutral current off metal parts, which is why the grounded conductor touches the case once, at the service disconnect, and nowhere after. Why the bond lives only at the service disconnect covers the rule, and keeping subpanel neutrals and grounds apart covers the wiring that follows it.
The old two-wire logic still clears faults, to be fair: a case bonded to the neutral bar returns fault current over the grounded conductor, so a short at that box trips inside one cycle like any copper path. The price arrives when a neutral breaks. With the feeder neutral open and no EGC, the bootleg bond makes the case the only return. Every neutral amp then flows through the box and the earth, the panel shows the divider symptoms above, and the case rides about 60 V above earth on the same 2.4 A path. Behind balanced loads the defect hides for decades, because half a volt on a dry floor is nothing anyone feels.
How do you check for an open neutral and a missing bond?
Confirm the open neutral in two meter readings: a steady line-to-line voltage at every outlet while leg-to-neutral readings swing from near 0 to nearly the whole supply, then shift when you switch a load from one leg to the other.
- Read the hots first: a steady 240 V between L1 and L2 means the transformer and both hots are healthy; the fault is in the return.
- Split the system: a steady nominal leg-to-neutral reading at the service disconnect with wild numbers in a subpanel puts the open in that feeder. Wild numbers at the main put it in the utility drop, which is the utility's conductor, so call them before pulling a single wire.
- Hunt a bootleg bond: with the panel dead and the feeder neutral lifted at its service-disconnect terminal, ohmmeter from that conductor to the subpanel enclosure. Megohms is clean; a few ohms or less is the stray bond.
- Prove the bond you want: neutral bar to enclosure at the service disconnect reads near 0 Ω, and anything more across a short jumper means corrosion or a loose lug.
- After repair, load it: run a 5 A load on one leg and read each leg to neutral. Anything outside 114-126 V, the ANSI C84.1 range-A band for a 120 V system, says the return is still weak.
How do you keep the fault path intact for the next 20 years?
Bond every metal enclosure, size every EGC from its breaker, and keep the bond count at one: a 20 A branch takes a 12 AWG copper EGC per NEC Table 250.122, a 100 A feeder takes 8 AWG, and the only legal neutral-to-case bond lives at the service disconnect.
That bond does its work only while the metal between panels can carry the current. NEC 250.96(A) requires metal parts to be bonded so they can safely conduct the fault current likely to be imposed on them, and NEC 250.12 strips paint and other nonconductive coatings off the contact surfaces at bonding connections unless the fitting is designed to bite through. Bare stainless makes this easy: the walls of a 20 x 20 x 8 in 304 stainless wall-mount enclosure — manufacturer-rated NEMA 4X / IP66, not UL listed — take a bonding lug straight onto clean metal. A painted steel box makes you strip a window in the coating under every lug, and washdown repaints close it again. Sizing the EGC from the service disconnect to a subpanel and bonding a disconnect enclosure to the building ground ring cover the two joints that corrode first.
When do you re-specify the enclosure instead of repairing it?
Re-specify when the fault path is the enclosure itself: a painted box with corroded bond points, a nonmetallic box with no EGC, or any enclosure whose bonding hardware cannot return to near 0 Ω. Replace it with a bare stainless box and fresh copper.
Then work the fix by case:
- Utility drop open, panel bonded and EGC'd: metal sits near 60 V and light loads run up to 218 V, so shut the main while people work nearby, call the utility, and re-check each leg under load after the splice.
- Feeder neutral open downstream of the bond: symptoms show only past the break, and any fault still clears in one cycle; repair the feeder, then confirm 114-126 V leg-to-neutral under a 5 A load.
- No EGC or a bootleg bond: take the circuit out of service, run the EGC, and move any neutral-to-case bond back to the service disconnect.
The hardware is the easy part: 304 or 316L stainless wall-mount enclosures cover the stock sizes, and a custom enclosure built to your cutout drawing covers the rest, delivered in 10 business days from your order or your drawing approval.
Enclosures in this article
| Enclosure | Material | Outside H x W x D / display | Rating |
|---|---|---|---|
| 20 × 20 × 8 in | 304 stainless | 20 x 20 x 8 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
Questions
Can a ground rod carry the fault current while the service neutral is open?
It closes the loop but far too weakly to trip anything. Two 25 Ω electrodes limit a fault at nominal leg voltage to about 2.4 A, 12% of a 20 A breaker's rating, so the breaker holds while the bonded enclosure sits near 60 V. NEC 250.4(A)(5) keeps the earth out of the effective ground-fault current path for exactly this reason: the copper grounded conductor does the returning, and the rod only anchors the system to earth.
How much voltage can a load see when the service neutral opens?
Anywhere from near 0 to the full 240 V line-to-line, set by the load split. Two legs at 48 Ω and 480 Ω divide it into about 22 V and 218 V, and the hot leg's electronics fail within seconds. Equal loads split it evenly, which is why the symptom appears and disappears as motors start and stop.
How fast does the breaker clear a fault when the return path is all copper?
In about one cycle. With roughly 0.5 Ω of copper in the loop, a fault at nominal leg voltage draws about 240 A through a 20 A breaker, twelve times its rating, and the magnetic element clears it in about 17 ms at 60 Hz. The same fault returned through two 25 Ω electrodes draws 2.4 A and holds forever.
Where is bonding the neutral to the enclosure legal?
At the service disconnect, once. The bond there lets fault current reach the grounded conductor; every panel downstream keeps neutrals on an isolated bar and grounds on the ground bus, because NEC 250.6 aims to keep neutral current off metal parts. A bootleg bond in a subpanel leaves 10 A of imbalance putting about 0.5 V on the case, and hands it the whole return when a neutral breaks.
