Subpanel neutral and ground bonding: keep them apart

Short answer: yes, the reversal was a real problem, and no, a subpanel's neutral and ground never tie together the way they do at the main. Bond them at the service disconnect only. In a subpanel, the neutral bar sits on insulated standoffs with the bonding screw out, and every amp of load current returns on the insulated neutral conductor. A Class A GFCI trips at 4-6 mA of neutral-to-ground current, and a bonded or swapped subpanel pushes amps through the grounding system instead. The fix is mechanical: white to the isolated bar, green or bare to the bonded bar, screw out, then a clamp check that should read under 0.1 A on the feeder ground.

Why does a bonded or swapped subpanel trip AFCIs and GFCIs?

Because load current returns through the grounding system instead of the insulated neutral: that shared metal path flutters and arcs at painted knockouts and loose setscrews, which is the exact signature an AFCI hunts for, and any neutral-to-ground contact past about 5 mA trips a Class A GFCI.

A GFCI compares the current in its hot leg with the current in its neutral and trips when the two differ. A US Class A device trips at 4-6 mA; the 30 mA level common on IEC RCDs is a different, higher protection point. With neutral and ground effectively tied at the subpanel, part of the load current wanders onto the grounding system, the breaker sees the difference, and it trips.

An AFCI watches the waveform for arc signatures. The grounding path runs through locknuts, setscrews and knockouts. Paint on a contact surface makes a poor joint: NEC 250.12 requires paint, lacquer and enamel removed from threads and contact surfaces at grounding and bonding connections. Amps flickering through that joint look, to an AFCI, like the arcs it exists to catch.

Here is the arithmetic for one panel. Take a subpanel 40 ft from the main, fed at 60 A on 6 AWG copper (65 A ampacity at 75 C) with a 10 AWG copper EGC, which NEC Table 250.122 sizes from the 60 A breaker. A 1500 W heater draws 1500 W / 120 V = 12.5 A. With the bonding screw in, the return splits between the neutral (40 ft of 6 AWG at about 0.4 Ω per 1000 ft = 0.016 Ω) and the ground path (40 ft of 10 AWG at about 1.0 Ω per 1000 ft = 0.04 Ω, with conduit in parallel ignored). The ground share is 0.016 / (0.016 + 0.04) = 29 %, so 12.5 A × 0.29 = 3.6 A rides the EGC, the conduit, the cabinet and every setscrew between them.

That 3.6 A through a joint carrying 0.5 Ω of contact resistance dissipates 3.6² × 0.5 = 6.5 W at one point, enough to cook paint and loosen the joint further. NEC 250.6 exists 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. The grounding and bonding guide covers the fault-path side in depth.

Should neutral and ground be tied together in a subpanel?

No — tie them together at the service disconnect and nowhere else. The NEC places the grounded-conductor-to-grounding-system bond at the service disconnect and bars that conductor from grounding any enclosure or equipment downstream, so a subpanel's neutral bar stays isolated.

The NEC puts the bond at the service: the grounded conductor bonds to the grounding system at the main disconnect and nowhere after it. The same rules carry that downstream: no grounded conductor grounds an enclosure past the service. Every panel fed from the load side of the main is a subpanel, whatever its label says.

The reason is path discipline. The earth is not an effective ground-fault current path under NEC 250.4(A)(5), and the grounding system is built for faults, not for load. Two bonds hand load current a choice of roads; the metal roads heat, corrode and flicker. With the feeder conductors reversed, the cabinet and its ground bar actually ride the neutral, and one loose neutral lug takes the fault path down with it — the same failure shape that turns a transformer insulation failure into a burned winding. Where the bond sits relative to the disconnect is its own question in metering too: see CT metering enclosure bonding before the service disconnect.

What symptoms point to a neutral-ground problem at the panel?

Watch for AFCIs or GFCIs that trip with no load fault, lugs or conduit fittings on the grounding path that run warm, and a clamp meter that reads amps on the feeder ground where a healthy installation reads milliamps.

Match what you see to where the cause likely sits:

• Several different AFCI circuits trip, no load fault found — the cause sits at the panel: a bonding screw left in, or a feeder on the wrong bars.
• One GFCI trips over and over — a neutral-to-ground contact on that one circuit, downstream of the breaker.
• Fittings, lugs or the ground bar run hot, or paint is scorched at a knockout — continuous load current is riding the grounding path. Metal you cannot hold a finger on is over about 50 C (122 F).
• A tingle from a bonded case, or a reading between metal and the concrete floor — neutral current on metal parts, which NEC 250.6 is written to prevent.
• A breaker tripping on a real overload is not a symptom — clear the load or upsize within the circuit's ratings before you blame the bars.

What causes neutral current on the ground wire?

Start with the bonding screw: convertible load centers ship with the green neutral bonding screw or strap installed, and leaving it in a subpanel is the most common cause; the feeder landed on the wrong bars comes second.

Work the list top down; the first two cover most panels:

1. The bonding screw or strap is still in the neutral bar. Convertible panels ship with it fitted for service duty. Look for a green screw in the bar or a metal strap from bar to cabinet.
2. The feeder is landed on the wrong bars. The white grounded conductor sits on the cabinet-bonded ground bar and the green or bare EGC sits on the isolated bar: the full reversal, with each amp returning through the grounding system.
3. One branch conductor is mis-landed. A single white on the ground bar, or a single bare on the neutral bar, puts one circuit's return on the wrong road.
4. A neutral touches ground downstream of a breaker. A staple through a cable, a conductor pinched under a device strap, a stray bare wire in a metal box: a few milliamps leak, and a Class A GFCI trips at 4-6 mA.

How do you check each cause?

Measure before you open anything: clamp the feeder's ground conductor with the loads running, read milliamps as healthy and treat 0.5 A or more as proof that load current is riding the grounding system.

In the worked example above, a bonded bar reads 3.6 A; leakage from filters and such stays in the milliamps. So 0.5 A is the working line: below it, look at the circuits; above it, look at the bars. Then open the dead front. Kill the feeder, lock it out and verify absence of voltage before any lug work; the lugs sit at line voltage even with a subpanel breaker off. With the panel open:

• Bonding screw: find it. In or out is a ten-second answer.
• Neutral bar isolation: the bar sits on plastic standoffs, and a meter from bar to cabinet reads open, in the megohms. A ground bar to cabinet reads about 0 Ω, under 0.1 Ω.
• Feeder landing: trace white to the isolated bar and green or bare to the bonded bar by conductor, never by bar position.
• One bad circuit: with its neutral lifted at the subpanel and its loads unplugged, measure from the lifted white to the ground bar. A near-0 Ω reading finds the downstream contact.

What is the fix for a bonded or swapped subpanel?

Put the feeder's white grounded conductor on the isolated neutral bar, put its green or bare EGC on the cabinet-bonded ground bar, and take the bonding screw out of the neutral bar — tape it inside the door so a future service conversion can still find it.

1. Keep the feeder locked out from the checks above. Move the feeder's white grounded conductor to the isolated neutral bar and its green or bare EGC to the ground bar.
2. Remove the bonding screw or strap from the neutral bar and tape it inside the door.
3. Re-land the branches: whites on the neutral bar, EGCs on the ground bar. An insulated EGC of 6 AWG or smaller is green or green with a yellow stripe under NEC 250.119.
4. Torque every lug to the value on the door label, commonly 20-45 lb-in on feeder lugs; an undertorqued neutral lug re-creates the flicker you just removed.
5. If the EGC is undersized, size it from the breaker ahead of it: a 60 A feeder takes 10 AWG copper and a 100 A feeder takes 8 AWG, per NEC Table 250.122.
6. Restore power, run the loads and clamp the feeder ground again: expect under 0.1 A, not amps. The AFCIs should hold from the first reset.

The ground bar itself bolts metal-to-metal to the cabinet wall. If you are adding or moving one, bonding an enclosure to a ground bus and ground bar placement on the wall vs the mounting plate cover the how and the where.

How do you keep the problem from coming back?

Label the panel as a subpanel, tape the bonding screw in the gutter, and repeat the clamp test after every job in the cabinet: the feeder ground should still read under 0.1 A with the loads running.

Write it where the next person reads it: NEUTRAL ISOLATED — BOND AT THE SERVICE DISCONNECT ONLY, inside the door. Then:

• Demoted panels: when a load center moves from service duty to subpanel duty, pulling the bond is step one, before any load lands.
• After any panel job: repeat the two-minute clamp test; it beats an hour of AFCI archaeology. Under 0.1 A on the feeder ground passes.
• Feeder changes: every re-termination lands white to the isolated bar first, EGC to the bonded bar second, then torque to the label.

When does the panel enclosure itself need a second look?

If the cabinet sees washdown, salt air or nightly condensation, corroded lugs and bar screws become the arcing joints that trip AFCIs, so move the panel into a 316L stainless enclosure rated NEMA 4X / IP66 by the manufacturer; not UL listed.

Water and corrosion are what turn good joints into marginal ones. Choose the grade by the chemistry in the room: 304 for most washdown and outdoor sites, 316L for salt, brine and chlorides. The NEMA enclosure types guide sorts rain, hose-down and dust by type. Any breather, drain or fan you add carries its own rating, and that rating becomes the assembly's rating — read what a breather vent does to a NEMA 4X rating before you cut one in. If the panel ends up inside a listed industrial control panel, sort the enclosure question first with enclosure listing for a control panel.

We build these to your drawing from one piece: custom stainless enclosures take your material, size and every cutout, and the 3D enclosure builder gets you a drawing to approve, delivered in 10 business days from that approval. Send the room's worst day with the quote request and the cabinet comes out right the first time.

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.

Questions