Short answer: a metallic gland is bare metal pressed against your conductors, so the day insulation fails onto it, the gland body goes live. Bonding the gland into the enclosure's earth path turns that fault into a trip: keep 0.1 Ω or less from gland body to the earth stud, and a 277 V fault pulls hundreds of amps and opens the breaker in one cycle. Leave the gland unbonded and the same fault can park the body at 272 V, waiting for a hand. The bond also gives surge and cable-screen currents a defined route to earth. The bond plays no part in sealing, which stays a separate check.
What the earth connection does on a fault
A brass or stainless gland grips the cable, but its body is also a metal part bridging your enclosure wall at the exact place the conductors pass. Faults reach it four ways: insulation nicked during the pull, a stray strand at a lug, abrasion inside the gland mouth, or water tracking along the cable. Any one of them puts a phase on the gland, and the gland, locknut, box wall and door all become part of the circuit.
The NEC states the rule directly: 250.96(A) requires metal parts that serve as the grounding path — raceways, cable armor, enclosures — to be bonded together so they conduct whatever fault current lands on them. Your return loop is gland, box wall, earth stud, equipment grounding conductor, source. Size that conductor from the rating of the overcurrent device ahead of it using Table 250.122, which sets 8 AWG copper for a 100 A device. NEC 250.4(A)(5) then closes the trap: the earth itself is not an effective path for clearing a ground fault, so the loop must be copper all the way.
The grounding and bonding guide covers the full scheme, and what to bond on a stainless box lists the parts, gland by gland.
Touch voltage and trip current: the arithmetic
Work a 480 V three-phase circuit through a metallic gland into a stainless box: 277 V phase-to-ground, a 20 A breaker, and 0.4 Ω of return loop through the equipment grounding conductor to the source. Only the gland joint changes between the two cases.
| Bonded gland at the target resistance | Unbonded, painted thread, 25 Ω | |
|---|---|---|
| Fault current | 277 V ÷ 0.5 Ω = 554 A | 277 V ÷ 25.4 Ω = 11 A |
| 20 A breaker | 28× rating: trips in ≈ 0.017 s, one 60 Hz cycle | 55% of rating: carries it indefinitely |
| Gland body voltage | 554 A across the joint ≈ 55 V for 17 ms | ≈ 272 V until someone finds it |
| Energy at the fault | ≈ 5,200 A²s let through | ≈ 3 kW burning at the thread |
Four rows tell the whole story. A bonded gland lives at 55 V for one cycle, brushing the 50 V AC that IEC 60364 treats as the limit for touchable metal, while the breaker clears it. An unbonded one sits at 272 V for hours, and the fault dissipates 277 V × 11 A ≈ 3 kW of heat exactly at the gland thread, inside your box.
The enemy is the high-impedance joint, not the imperfect one. Double the joint to 0.2 Ω and the fault still reaches 277 ÷ 0.6 = 462 A, twenty-three times the breaker rating, still an instantaneous trip. The 25 Ω painted thread is what kills.
The second job: surge and screen currents
Fault clearing is half the reason for the bond. The other half is current that arrives from outside and needs a way out.
A nearby lightning strike induces surge current in any cable run long enough to act as an antenna. That current reaches the gland and looks for a path: across 1 Ω of bond, a 10 kA, 8/20 µs surge develops 10 kV. Cut that bond impedance tenfold — the figure you are building to at the gland — and the panel sees ten times less voltage on its metalwork.
A screened cable does the same job continuously. Terminate the braid or armour in 360° contact around the gland entry — cable shield termination at the enclosure entry says what to clamp and what to leave floating — and RF noise stays off your signal wires; how much RF leaks through enclosure openings shows why the entry dominates. Bonding a screen at both ends invites circulating current that heats its own path: induced heating at cable entries covers the failure, and the arithmetic of bond strand count versus strap size in circulating loss tells you which strap to order.
The numbers that prove the bond
Three values cover the installation:
- Gland body to earth stud: 0.1 Ω or less. This is the joint you own — the washer, the tag, the stud.
- Any bonded metalwork to the earth terminal: under 1 Ω. Static-control and equipotential-bonding practice works to this figure for the whole panel.
- A driven rod: 25 Ω or less, and only as an addition. NEC 250.54 lets you connect an auxiliary electrode to the equipment grounding conductor, and at 25 Ω a 277 V fault pushes about 11 A into soil — the unbonded case again — so the rod supplements the copper loop rather than replacing it.
Measure with the circuit locked out and proven dead. A two-wire multimeter proves continuity exists, but its leads alone read 0.2 to 0.3 Ω, three times your target; zero them or use a four-wire micro-ohmmeter that injects 10 A. Probe the gland body itself rather than the tag washer you fitted.
Keep the tests separate, too. The hose test behind an IP66 marking measures water, and the bond meter measures ohms. A box rated NEMA 4X / IP66 by the manufacturer — not UL listed — proved its water and dust performance; the gland bond is yours to build and measure.
Where gland bonds fail: paint, sealant and threads
NEC 250.12 requires paint, lacquer and enamel removed from threads and contact surfaces at grounding connections, and the gland thread is where this bites. A powder-coated box with a gland torqued over the coating, or a thread sealed with silicone, is your 25 Ω case: sealed beautifully, bonded barely.
Build the joint deliberately:
- Bare metal at the hole: deburr through the coating so the locknut seats on steel.
- A serrated washer under the locknut, or a listed earth tag bridging gland body to box.
- The gland plate bonded back to the wall, because a bolted plate is a joint too — when to specify a gland plate helps you decide the layout.
- The door on a braid strap, since hinges carry current only by accident; see door bonding straps and continuity.
- Bond conductors of 6 AWG and smaller in green or green with a yellow stripe, per NEC 250.119, so the next person knows the wire's job.
What to specify
Put the bond on the drawing, because it is cheap there and expensive in the field:
| Item | What to order |
|---|---|
| Glands | Metallic, with locknut and earth tag for every armoured or screened cable |
| Contact surfaces | Gland holes machined through the coating; serrated washers in the hardware bag |
| Earth stud | One per backplate, sized for the Table 250.122 conductor |
| Bond conductor | Green or green with a yellow stripe, landed on the ground bus |
| Door and gland plate | Braid straps, fitted at assembly |
A 12 x 12 x 6 in [305 x 305 x 152 mm] 316L wall-mount enclosure takes a small motor-starter group with room for the stud and straps; an 8 x 8 x 4 in [203 x 203 x 102 mm] 316L box covers a junction duty with two or three glands. Send the gland-hole sizes and positions on the cutout drawing, pick the grade from the 316L wall-mount range for washdown duty, and the box is delivered in 10 business days from your approval of the drawing.
The decision in three lines: metallic gland on any powered cable — tag, serrated washer, bond the gland to the stud at the resistance figure above. Screened or armoured cable — 360° termination at the gland plus a strap sized for the circulating current. Non-metallic gland — no bond to make, so spend the time on the clamping range and the seal instead.
Enclosures in this article
| Enclosure | Material | Outside H x W x D / display | Rating |
|---|---|---|---|
| 12 × 12 × 6 in | 316L stainless | 12 x 12 x 6 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
| 8 × 8 × 4 in | 316L stainless | 8 x 8 x 4 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
Questions
What resistance should a gland-to-earth bond measure?
0.1 Ω or less from the gland body to the earth stud, measured with a zeroed meter or a four-wire bond tester. Panel-wide bonding to the earth terminal stays under 1 Ω. A two-wire multimeter reads its own leads at 0.2 to 0.3 Ω, so zero them first or the number means nothing.
Do non-metallic glands need earthing?
The gland itself needs no bond, because a polyamide body has no metal to go live. The cable screen or armour it carries still needs its own terminator bonded to the box, and the seal still needs attention: a plastic gland holds an IP66 entry only when its clamping range matches the cable outside diameter.
Is a locknut enough to earth a metallic gland?
On a clean, bare hole a torqued locknut can measure a fraction of an ohm and it holds there. Over paint, powder coat or sealant the same joint reads tens of ohms, and a fault rides at hundreds of volts. NEC 250.12 requires coatings removed at contact surfaces, so fit a serrated washer or an earth tag across the gland.
Does gland earthing change the IP66 or NEMA 4X rating?
No, the two tests never meet. The hose test behind an IP66 or 4X marking measures water ingress; the bond is electrical and measured in ohms. Check them as separate jobs: the gland-to-stud bond reading for the electrical side, and gasket condition, locknut torque and clamping range for the seal.
Does a ground rod at the enclosure earth the gland?
No. NEC 250.4(A)(5) is explicit that earth is not an effective ground-fault path, so at 25 Ω a 277 V fault pushes only about 11 A into soil and the breaker stays shut. Under NEC 250.54 a rod is an auxiliary electrode, connected alongside the equipment grounding conductor and never in place of it.
