Short answer: the manufacturer's arrangement stands. Nothing in the NEC forces a dedicated conductor from a transformer to the substation grid, and IEEE 80 builds substation grounding on exactly this kind of bonded structure. The condition is simple: every bolted joint between the cabinet and the grid must be a real bond — bare metal, proper hardware, able to carry fault current until a breaker clears. The utility engineer's worry is best practice, not code: bolts through painted, weathering steel drift over decades. Run the numbers and you will see the grid tie carries almost none of a secondary fault anyway, about 5 A of 1040 A, so you win the case on the joints — and a 6 AWG copper jumper ends the argument for a few feet of wire.
Two bonds with different jobs
A control-power transformer fed at 480 V and delivering 240/120 V from its own winding is a separately derived system: the secondary has no metallic tie back to the source, so its neutral takes its earth reference where you bond it. Three connections do three different jobs, and the argument with the utility is only ever about the third. The enclosure grounding and bonding guide covers the general case; this one turns on what each conductor is for.
| Connection | What it does | What the NEC sizes it at | Where it lands |
|---|---|---|---|
| System bonding jumper, X0 to cabinet | Carries the whole fault current when a secondary line hits the case | With 10 AWG copper secondary conductors, 8 AWG copper | The cabinet ground bus, at one point only |
| Equipment grounding conductors | Return a fault on each 120/240 V circuit so its breaker trips | From the breaker rating, NEC Table 250.122: 10 AWG copper covers 30 through 60 A devices | The same ground bus |
| Grounding electrode conductor | References the system to earth; carries almost no fault current | The electrode rules start at 8 AWG copper, and a rod, pipe or plate connection never needs more than 6 AWG copper or 4 AWG aluminum | The substation grid, directly or through bonded structure steel |
Land the load neutrals on an insulated neutral bar and the bonds on the ground bus. NEC 250.6 is what keeps unbalanced neutral current off dead metal: with a single X0 bond, that current returns on copper sized for it. Bond the neutral at two points and the current splits through the cabinet and the breaker frame — the arrangement the code calls objectionable.
Where the fault current really flows
Trace a fault from a 120 V secondary line to the cabinet before you argue about the grid. The current leaves the winding, touches the case, and comes home through the cabinet steel to the X0 bond and back to the transformer — a loop measured in milliohms, and that loop is what trips the breaker. The tie to the grid runs parallel to it through the earth, and NEC 250.4(A)(5) says plainly that the earth is never an effective ground-fault path. A 25 Ω earth loop — the resistance a single ground rod is held to under the NEC — passes 4.8 A, while the metal loop passes about 1000 A. Your secondary breaker clears without the grid's help.
The 480 V side works the same way: a primary fault returns on the primary feed's own equipment grounding conductor, back at the source that feeds it. A winding-to-core failure is a different failure mode again, and transformer insulation failure bonding walks through it.
What the codes accept
Read the separately derived rules as written and they ask one thing: bond every dead metal part to the grounding point of the derived system — transformer case, cabinet, breaker frame. NEC 250.96(A) lets metal enclosures and the structures supporting them serve as the equipment grounding path once they are bonded to carry the fault current likely to be imposed on them. A bolted structure path is a recognized method, in other words, not a way around one. NEC 250.12 then tells you what disqualifies a joint: paint, lacquer and enamel come off the threads and contact surfaces at bonding connections. A bolt riding on coating is the arrangement that fails the code test, and it hides in most cabinets grounded through the bolts alone.
IEEE 80, the substation grounding guide, builds the whole yard on this method: structures and even the fence bond to the grid, because during a 345 kV fault the grid and the bonded steel rise together, and touch voltage is the difference between two points. The engineer pressing for a dedicated conductor is defending that yard model against 30 years of corrosion at joints nobody re-inspects. Fair position — the NEC sets the floor, not the ceiling, and a utility's station criteria may sit above it.
The numbers for a 7.5 kVA, 480-240/120 V transformer
Work the transformer at the center of the dispute:
- Nameplate: 7.5 kVA, 480 V primary, 240/120 V single-phase secondary.
- Full-load secondary current: 7500 VA ÷ 240 V = 31.3 A, which takes 10 AWG copper conductors and a 35 A breaker.
- Nameplate impedance: small dry-type control transformers sit between 2 and 6 %; use 3 %.
- Magnetic pickup: molded-case breakers set it at roughly 5 to 10 times handle rating, 175 to 350 A here.
- Earth loop through the grid: 25 Ω, the value a single driven rod is held to under the NEC.
The rules in words: available fault current is full-load current divided by per-unit impedance; the breaker clears on its magnetic curve only while fault current sits above pickup; and earth-return current is fault voltage divided by earth-loop resistance.
The arithmetic:
- 7500 VA ÷ 240 V = 31.3 A full load.
- 31.3 A ÷ 0.03 = 1040 A available at the terminals; count the 480 V feed's impedance and you can plan on 1000 A at the cabinet.
- 1040 A sits 3 to 6 times above the 175-350 A pickup band, so the breaker clears on magnetic, in about 0.05 s — three cycles at 60 Hz.
- Loop budget: 120 V ÷ 350 A = 0.34 Ω. The steel-and-bond loop runs tens of milliohms; at 20 mΩ you sit 17 times inside the budget.
- Earth parallel: 25 Ω passes 4.8 A, about 0.5 % of the fault. The grid tie clears nothing.
The decision: hold the metallic loop under 0.34 Ω and a line-to-case fault dies in 0.05 s. Let one painted or corroded joint push the loop to 0.7 Ω — roughly double the budget — and the current falls to 171 A, below the pickup band: the breaker clears on thermal instead, in seconds, with the cabinet sitting near line voltage the whole time. One joint decides which afternoon you have.
Checks that prove the bolted path
Kill the 480 V feed and prove it dead, then earn the bond with your wrench and a meter:
| Check | What passes |
|---|---|
| Contact faces at each joint | Bare metal under the hardware, per NEC 250.12, with toothed washers or a ground boss |
| Whole path, cabinet to grid | Under 0.1 Ω on a 4-wire microhmmeter; single-digit milliohms per joint |
| The door | A bonding strap or spring contact, so it is dead metal like the rest — door bonding and continuity |
| The gland plate | Bonded to the cabinet, with glands earthed where armor lands — cable gland earthing and the fault path |
| Mixed metals outdoors | Stainless against galvanized steel creeps; isolate or match — stainless and galvanized contact corrosion |
Photograph each joint and file the readings. That record settles an objection faster than any citation, because it shows the path was measured rather than assumed.
The decision, case by case
If your joints are documented bonds — bare faces, bonding hardware, milliohm readings — keep the manufacturer's arrangement; it meets the NEC as written. Pull a 6 AWG copper jumper from the cabinet ground bus to the nearest grid riser anyway: a few feet of wire, inspectable forever, and the objection retires. Terminate it the way bonding an enclosure to a ground bus shows, on a proper lug over bare metal.
If the joints are painted or unreadable, pull one bolt and look. Re-making every joint with bonding hardware costs a day; pulling one conductor costs an hour, and the conductor wins on inspectability for the next 30 years.
If the utility's station standard insists on a dedicated conductor, run it. On their side of the fence their standard and the NESC are the rule, and the NEC only sets the floor.
Building a new control cabinet? Put the ground-bus cutouts, the gland plate and the bonding hardware on the drawing. You set the material, the size and every cutout on a custom enclosure, and it is delivered in 10 business days from your approval of the drawing.
Enclosures for this job
Catalog sizes that suit the job this article answers, picked from its subject; open one for its full specifications.
| Enclosure | Material | Outside H x W x D / display | Rating |
|---|---|---|---|
| 60 × 36 × 16 in | 304 stainless | 60 x 36 x 16 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
| 48 × 36 × 16 in | 304 stainless | 48 x 36 x 16 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
| 60 × 36 × 12 in | 304 stainless | 60 x 36 x 12 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
Questions
Does driving a ground rod at the cabinet fix the objection?
No, and it can quietly make things worse. Under NEC 250.54, an electrode added at an enclosure is an auxiliary electrode: you may tie it to the equipment grounding conductor, but it never replaces the metallic fault path, because a 25 Ω earth return passes about 5 A while the breaker needs hundreds to trip fast. The substation grid is already the electrode here; a second rod adds another reference, not a fault path.
What size conductor should the direct tie to the grid be?
Use 6 AWG copper, or 4 AWG aluminum if the run and the terminations suit it. The NEC's electrode rules never ask for more than that to a rod, pipe or plate, and the tables start at 8 AWG copper for a secondary wired in 10 AWG. Keep the run short and out of ferrous conduit: a lone conductor inside steel conduit gains inductive impedance unless the conduit is bonded at both ends, and that extra impedance shows up exactly when you need the path low.
Why does the utility care if the grid tie carries almost no current?
Because a 345 kV fault puts kiloamps through the yard grid, not 5 A. For example, 10 kA into a 0.5 Ω grid lifts every bonded structure 5000 V together, and touch safety comes from the steel rising as one piece; corrosion at a single joint breaks that model. The engineer is defending a yard design with 30-year assumptions against joints nobody re-inspects, and the fault current from one small control transformer never tests it.
How do you measure the bond path so both sides accept it?
Use a 4-wire microhmmeter or a ground-bond tester that pushes at least 10 A through the path; a multimeter with 0.1 Ω resolution cannot tell a sound joint from a painted one. Expect single-digit milliohms per bolted joint and under 0.1 Ω from cabinet to grid. Measure with the door closed as well as open, record the instrument and the readings, and attach both to the commissioning file.
Does the 480 V primary side need its own grounding conductor?
Yes. The primary feed carries an equipment grounding conductor sized from its overcurrent device under NEC Table 250.122: 12 AWG copper for the 20 A breaker a 7.5 kVA unit takes at 480 V (7500 VA ÷ 480 V = 15.6 A full load). It runs with the primary conductors back to the station service source, and a primary fault clears through that conductor, not through the yard grid.
