Short answer: a metal nipple or raceway between two panels carries neutral current only when the neutral is bonded to the enclosures at both ends. Keep the neutral isolated at the subpanel and the steel path stays what the code wants, a bonding route at near 0 A. Bond the neutral in both boxes and the raceway becomes a second neutral conductor. The worked example below sends a 30 A unbalance and puts about 21 A through a 3 in [75 mm] nipple and its four locknut joints, with the arithmetic on the page.
When does a metal raceway between two panels carry neutral current?
Only when the neutral touches metal at both ends. With the neutral bonded inside the main panel and again inside the subpanel, the return current sees two parallel paths: the copper wire and the steel wall. The amps divide between them.
Two nodes run the circuit. The subpanel neutral bar is one, the main panel's neutral bus is the other, and every amp that returns must travel from the first to the second. Path one is the neutral conductor itself. Path two exists only when a bond screw or strap ties the neutral bar to the subpanel wall. From there the current crosses the nipple into the main panel wall and back to the bus through the main bonding jumper. Parallel paths share current in inverse proportion to impedance, so the lower-impedance path takes the larger share.
The steel path is always present, because the nipple is bonded at both ends for fault duty; that part is correct and required. The bond screw in the subpanel is the only connection that hands load current to it. This is the same one-end principle that protects single-core armoured cable, where a gland bonded at both ends invites circulating current around the sheath; the page on single-core armoured cable glands and one-end bonding walks through that case. For the panel itself, the article on subpanel neutral and ground bonding covers the separation in detail.
How much neutral current does the nipple actually pick up?
Follow the impedances: in the worked example a 30 A unbalance on a 6 AWG neutral puts 21.4 A through the raceway path and leaves 8.6 A on the copper, because the bonded steel path sits near 2 mOhm against 5 mOhm of wire.
Assume a 120 V branch pulling 30 A on one leg of a 240/120 V feed, so the neutral carries the full 30 A back. The feeder is protected at 60 A and the neutral is 6 AWG copper, good for 65 A at 75 C under the NEC ampacity table for copper conductors. The effective bus-to-bus length is 10 ft. The NEC resistance tables list 6 AWG copper at about 0.49 ohm per 1,000 ft at 75 C, so the wire path is 0.49 x 10 / 1,000 = 4.9 mOhm, call it 5 mOhm. Assume the steel path (two enclosure walls, four locknut and thread joints, 3 in [75 mm] of threaded nipple) totals 2 mOhm with the joints tight. Those two figures are the assumptions; replace them with your own lengths and joint count.
Ohm's law for parallel paths sets the split: I(raceway) = 30 A x 5 / (5 + 2) = 21.4 A, and the neutral conductor carries 30 - 21.4 = 8.6 A. More than two thirds of the return current leaves the wire the feeder schedule paid for and rides the panel steel.
| Return path | Assumed impedance | Share | Current |
|---|---|---|---|
| Steel: two walls, four joints, 3 in nipple | 2 mOhm | 71% | 21.4 A |
| Copper: 10 ft of 6 AWG | 5 mOhm | 29% | 8.6 A |
Joint quality moves the split. Let the steel path corrode toward 5 mOhm and the split evens out at 15 A and 15 A. Polish it toward 1 mOhm and the steel takes 25 A of the 30. Near this operating point every added milliohm of contact resistance moves roughly 3 to 4 A from the steel back to the copper. Heat follows the square of current, so drop the unbalance to 15 A and the steel carries 10.7 A while the joint watts fall to 1.1 W, a quarter of before.
Why is a bonded raceway a heating and arcing hazard at loose joints?
Those 21.4 A cross four mechanical joints continuously. One loose locknut at 10 mOhm dissipates 21.4 squared x 0.01 = 4.6 W in a washer-sized spot, enough to cook finish and gasket. An intermittent joint strikes an arc each time it opens under load.
Locknuts loosen. Vibration, thermal cycling and a joint that already runs hot all back the contact pressure off, and resistance climbs as the true contact area shrinks. The watts concentrate into that shrinking area, the wall beneath discolors, and the next fault has to cross a path that daily duty has been burning. NEC 250.4(A)(5) calls the raceway part of the effective ground-fault current path, the low-impedance route that lets the breaker clear. A nipple joint carrying 21 A of load current every day is exactly the spot that duty degrades.
The other hazard is arcing. An intermittent contact carrying 21 A arcs on every make and break, and an arc inside a panel spits molten metal at whatever sits in front of it. A joint carrying near 0 A, the isolated-subpanel case in the next sections, has neither the heat nor the arc duty.
What does the NEC require for a metal nipple between panels?
Bond the nipple at both ends and let it carry faults, not load. NEC 250.96(A) requires raceways and enclosures serving as equipment grounding conductors to be bonded so they can safely conduct any fault current imposed on them, and nothing in that duty includes neutral amps.
Three rules set the design. First, the NEC requires all conductors of one circuit, the grounded conductor included, to run in the same raceway, so the neutral rides in the nipple with the hots; the bonding of a metal junction box with spliced feeder conductors covers the landing details. Second, the bonded nipple can serve as the equipment grounding conductor on its own, or it can run parallel to a wire sized from NEC Table 250.122, which asks for 10 AWG copper ahead of a 60 A breaker. Third, the NEC requires ferrous metal raceways and enclosures around grounding conductors to be bonded at each end; an unbonded steel wall around current-carrying conductors heats and induces. The enclosure grounding and bonding guide collects the rest of these rules.
How do you keep the raceway at near 0 A under load?
Land the neutral once. In the subpanel the neutrals terminate on an isolated block with no metallic connection to the box, the equipment grounds land on the ground bus bonded to the wall, and the single neutral-to-case bond lives at the main panel or service equipment.
NEC 250.6 is written for exactly this: it aims to keep neutral current off metal parts, which is why neutrals land on an isolated terminal and grounds on the ground bus. In practice you do four things. Pull the factory bond screw or strap that ties the subpanel neutral bar to the case. Mount the ground bus on the bare wall and land every green conductor and nipple ring there. Confirm the neutral bar stands on insulated standoffs, so its only connection home is the neutral conductor. Leave the bonding point upstream, where the main bonding jumper meets the service neutral.
Done this way the nipple carries a load-current duty of 0 A and saves itself for faults. It becomes a bonded enclosure path with nothing to heat and nothing to arc, which is the outcome the whole subpanel separation exists to reach.
What happens if a neutral-ground bond stays in the subpanel?
The stray bond hands the return current its second connection. The same 30 A unbalance starts moving about 21 A through the nipple on the first day, and the signs are warm locknuts, discolored rings around the joints and 60 Hz hum on nearby signal cables.
Watch for three things once the panels run an unbalanced load:
- Warm metal at the joints. A locknut point dissipating 4.6 W sits clearly above ambient; bare-hand warmth at moderate load is the classic sign.
- Discoloration. Burnish or dull rings on the wall around each locknut, growing over weeks of duty.
- Hum in analog wiring. A steel wall carrying 21 A induces 60 Hz into signal cables run through the same nipple, the effect measured in the article on running parallel conductors through a steel enclosure without induction.
The fix is mechanical and cheap. De-energize the feeder, lock out and verify absence of voltage, open the subpanel and pull the bond screw or strap joining the neutral bar to the case; the walk-through for removing a neutral ground bond in an outdoor subpanel covers the steps and the re-test. Confirm the neutral bar stands on insulators, reland, restore power and run the unbalanced load again. The nipple should stay at room temperature, and the voltage check in the FAQ below verifies it in ten seconds.
What should you order for a nipple-fed subpanel pair?
Order two gasketed stainless wall-mount boxes sized for the breakers, a concentric hole pair cut for the nipple, an isolated neutral block plus a wall-bonded ground bus in the subpanel, and bare locknut contact on the stainless walls. Both boxes are manufacturer-rated NEMA 4X / IP66 and not UL listed.
A six-to-eight-circuit subpanel fits a 20 x 16 x 8 in [508 x 406 x 203 mm] 304 stainless wall-mount enclosure with room for handle swing and the nipple landing; the 316L version is the pick for salt, brine and washdown. Inside depth is 7.88 in [200 mm], which leaves space behind the nipple. Mount so the highest breaker handle stays no more than 2.0 m (6 ft 7 in) above the floor per NEC 240.24(A), and keep 3 ft (900 mm) of clear depth in front for 0-150 V-to-ground gear under NEC 110.26(A), with the door swinging a full 90 degrees. Wall-mount sizes run from 6 x 4 x 4 in up to 60 x 36 x 16 in, so the wall-mount enclosure range covers both panels of the pair.
Order the two holes concentric and matched to the nipple trade size as custom enclosures with your own cutouts; they arrive delivered in 10 business days, counted from your approval of the drawing. Build the joint as the article on extending a junction box off a panel with a conduit nipple shows. On any powder-coated box scrape the coating from the locknut contact faces: NEC 250.12 requires paint, lacquer and enamel removed from threads and contact surfaces at grounding and bonding connections unless the fitting is designed to bite through. Brushed stainless walls need no scraping.
Catalog dimensions to compare
Compare these catalog dimensions with your component layout. Open each enclosure for its rating, material and usable space, then check those specifications against your application.
| Enclosure | Material | Outside H x W x D / display | Rating |
|---|---|---|---|
| 20 × 16 × 8 in | 304 stainless | 20 x 16 x 8 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
| 20 × 16 × 8 in | 316L stainless | 20 x 16 x 8 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
Questions
How many amps can a conduit nipple carry as a neutral?
A bonded nipple is a defect rather than a designed conductor, so it carries no ampacity listing to look up. The worked example puts 21.4 A of a 30 A unbalance through an assumed 2 mOhm steel path while the 6 AWG neutral carries 8.6 A. Read those amps as evidence of a stray bond, remove the bond at the subpanel, and the steel returns to fault-only duty.
Does a ground rod at the subpanel stop parallel neutral current?
Almost none of it moves to the earth. Two 25 ohm electrodes in series give the soil path about 50 ohms against a nipple path near 2 mOhm, so the ground carries about 0.004 percent of a 30 A unbalance. NEC 250.54 also classes a subpanel ground rod as an auxiliary electrode, never a substitute for the equipment grounding conductor.
What should a meter read between neutral and enclosure at a correct subpanel?
Measure under a real load unbalance, not at idle. Neutral bar to enclosure shows the wire drop alone: about 0.15 V for 30 A on the example's 10 ft of 6 AWG (30 A times 5 mOhm). A dead 0.00 V while load runs means the neutral is bonded to the box, so pull the bond screw or strap.
Can a plastic nipple between the panels end the problem instead?
It ends the parallel path cleanly: a non-metallic nipple carries no current at any load. The feeder then needs a wire equipment grounding conductor sized from NEC Table 250.122, 10 AWG copper for the 60 A feeder in the example, and both panels still get bonded, with the neutral isolated in the subpanel exactly as before.
Does the nipple bond still matter once the neutral is isolated?
Yes, with a changed duty. Day to day the nipple sits near 0 A, then a ground fault hands it the full fault current to carry back and trip the breaker, which is why NEC 250.4(A)(5) treats it as part of the effective ground-fault current path. Keep the locknuts tight and both ends bonded for that moment.
