What a Ground Fault Is, as Distinct From an Overload

Why this matters

A customer says they get a tingle off a metal cabinet when the floor is wet. The tech checks the panel, sees the breaker holding, and reports the circuit sound. Both of those observations are accurate and the conclusion is dangerous, because the breaker was never able to see what the customer felt. The three protective questions on an ordinary branch circuit live about three orders of magnitude apart, and a device only ever answers the one it was built for.

If anyone reports a shock or a tingle from equipment or a metal surface, that is the first action, not a diagnosis: keep everyone off the surface and out of any standing water around it, de-energize the circuit at the panel, and prove dead at the equipment with a tester checked live-dead-live per NFPA 70E-2021 120.5 before touching anything. An energized enclosure that has not tripped anything will stay energized indefinitely, which is exactly what makes it worse than a fault that opened a device.

Three current scales, one circuit

Nothing spans those scales. A 20 A thermal element has no mechanism for noticing a few milliamps, in the same way a truck scale has no mechanism for weighing a letter. That is not a defect in the device and no amount of resizing fixes it.

There is a fourth scale on larger services: ground-fault protection of equipment, which watches for ground-fault current at levels far above what hurts a person but below what an overcurrent device would clear promptly, and which the adopted NEC requires on solidly grounded wye services above a defined voltage and ampere threshold in Article 230, with a maximum setting and a maximum clearing time stated there. It protects the equipment and the building, not people. Read that article in the edition your authority having jurisdiction has adopted rather than carrying a remembered number to a job.

The distinction that actually separates them: the return path

An overload is too much current on the intended path. It goes out on the ungrounded conductor, through the load, and back on the grounded conductor, and both conductors carry the same magnitude. Nothing is misrouted; there is simply more of it than the wire is rated for.

A ground fault is current returning outside the intended path. It leaves on the ungrounded conductor and comes back through metal, earth, water or a person instead of through the grounded conductor. What defines it is not the size, it is the route.

 source ---- ungrounded conductor ----> load ---+
   |                                            |
   |<--- grounded conductor, normal return -----+
   |                                            |
   |                                    metal enclosure
   |                                            |
   |<--- equipment grounding conductor ---------+
   |     low impedance, large current,
   |     the overcurrent device clears it fast
   |
   +<--- person, wet floor, corroded path ------+
         high impedance, small current,
         the overcurrent device never moves

The bottom two paths are the same fault at the enclosure. What differs is the impedance of the route home, and that difference decides which device, if any, ever finds out.

Why the overcurrent device is sometimes the right ground-fault protection and sometimes blind

This is the part that gets stated in only one direction, so state both ends.

Low-impedance path. When the fault lands on metal that is bonded back to the source through an effective ground-fault current path, the loop impedance is small and the current is large - hundreds or thousands of amps. That is above the magnetic element's pickup, commonly around ten times device rating on a standard molded-case breaker though the product curve governs, and the device clears in a fraction of a cycle. This is the intended design: the grounding and bonding system exists to make ground faults big enough to be seen by ordinary overcurrent protection. NEC Article 250 in the adopted edition governs how that path is established, and the sibling articles on grounding and bonding cover it properly; do not re-derive it from here.

High-impedance path. When the return is through a wet floor, a corroded joint, a painted surface or a person, the loop impedance is large and the current is small - milliamps to a few amps. Nothing in the overcurrent device responds. The metal stays energized and the circuit keeps working normally, which is why nobody reports it as a fault. They report it as a tingle.

That is the whole reason a separate class of device exists. The dangerous band sits above what a person tolerates and below what a conductor cares about, and no device sized for the conductor can be made to cover it.

Field symptoms that separate the two before you open anything

• Overload trips are load-correlated and slow. They arrive minutes into a run, they track with what is switched on, and they repeat at roughly the same elapsed time.
• Ground-fault trips are condition-correlated and can be instant. They track with wet weather, wash-down, a specific piece of equipment being energized, or a physical disturbance of a cable, and they can happen at any load including nearly none.
• A ground-fault device that trips with everything unplugged is reporting the wiring, not the load. The most common finding in that case is contact between the grounded conductor and grounding metal somewhere downstream of the device, which the sibling article on why neutral and ground are not the same explains.
• An overload trip on a circuit at a small fraction of its rating is a contradiction. Either the current is not what you think, the ambient at the device is doing the work, or you are not looking at an overload at all.

Worked example: the tingle nobody could find

A 20 A branch circuit feeding a piece of equipment in a wash-down area. Intermittent tingle from the enclosure, worse when the floor is wet. No trips, ever. Two previous visits found nothing.

Following the safety action above first: the circuit is de-energized at the panel, locked out under 29 CFR 1910.333(b)(2) in general industry or 29 CFR 1926.417 on construction work, since 29 CFR 1910.147 excludes exposure to electrical hazards from work on conductors and equipment in electric utilization installations at (a)(1)(ii)(C), and proved dead live-dead-live per NFPA 70E-2021 120.5. Any live reading taken afterwards falls under 29 CFR 1910.333(a)(1), which requires de-energizing unless the employer can demonstrate that it introduces additional or increased hazards or is infeasible, with the boundary and arc-rated PPE from the risk assessment at NFPA 70E-2021 130.5 and 130.7 as adopted by the employer's program. A differential leakage reading only exists with the circuit running, which is the infeasibility the employer's program has to have already addressed in writing.

The measurement that finds it is a clamp placed around the ungrounded and grounded conductors together, so the meter reads the difference between what left and what came back. That difference is by definition the current going home some other way.

• Baseline, equipment running dry: 3.2 mA of difference.
• Same equipment, floor wetted as it is in normal service: 5.8 mA.

Put 5.8 mA against the three scales stated above.

• Against the 20 A device: 20 A divided by 5.8 mA is a factor of about 3,400. The circuit would need roughly three and a half thousand times that current before the thermal element noticed anything, and the conductors would be long gone first. The breaker is not failing to report. It has nothing to report.
• Against a Class A ground-fault device's nominal 4 to 6 mA band: 5.8 mA is inside it. A ground-fault device on this circuit would be operating, probably intermittently, exactly as the wet floor came and went.

So the finding is a real insulation failure in the equipment, sized in milliamps, on a circuit whose only protection is sized in tens of amps. De-energized and disconnected from the supply, an insulation resistance test between the ungrounded conductor and the enclosure confirms it. That test is an instruction that creates its own hazard: it applies hundreds of volts DC and charges cable capacitance, so disconnect any electronics that cannot take the test voltage, keep people off the far end, and discharge through the instrument and confirm zero before touching the conductors.

Now run the other end of the range on the same fault. Had the same degraded insulation made solid contact with the bonded metal enclosure rather than leaking through moisture, the loop impedance would have been small, the current would have been in the hundreds or thousands of amps, it would have passed the magnetic element's roughly ten-times-rating pickup, and the breaker would have opened in a fraction of a cycle on the first day. Identical failure, identical equipment, opposite outcome, and the variable is the impedance of the return path.

The failure mode of getting this wrong: the tech reports the circuit sound because the breaker held, or worse, adds a ground-fault device and calls the resulting trips a nuisance. The device would be doing its job. The insulation is still failing and the enclosure is still energized whenever the floor is wet.

What would change the response: if the differential reading were near zero with the equipment running and the tingle persisted, the source is not this circuit, and the next candidates are a neighbouring circuit sharing metal, a lost or high-impedance bonding path back to the source, or a voltage difference between two separately grounded metal systems. Those are grounding and bonding questions and belong to the articles that own them.

How to verify you got this right

• Name the return path before you name the device. If you cannot say where the current is going home, you have not identified a ground fault, you have identified a trip.
• Check that your measurement is a difference, not a magnitude. A clamp around one conductor measures load. Around both, it measures leakage. Reading the wrong one is why previous visits found nothing.
• Confirm the scale matches the device you are blaming. Write the ratio down. If the leakage is thousands of times below the overcurrent device's rating, that device is not part of the story in either direction.
• Re-test after the repair with the condition present, not with the equipment dry and idle. A wash-down fault that only appears wet is only proved fixed wet.

References

• NEC Articles 210, 230 and 250, in the edition your authority having jurisdiction has adopted
• UL 943, the listing standard for ground-fault circuit interrupters, in the edition the installed device was listed under
• 29 CFR 1910.333(a)(1) and (b)(2); 29 CFR 1910.147(a)(1)(ii)(C); 29 CFR 1926.417 for construction work
• NFPA 70E-2021, 120.5, 130.5 and 130.7, as adopted by your employer's electrical safety program
• See related: What a Ground Fault Actually Is; Why Neutral and Ground Are Not the Same; Grounding and Bonding as Two Different Jobs