What a Ground Fault Actually Is

Why this matters

The most dangerous electrical fault a field tech meets is not the one that makes a bang. It is the one that quietly puts a metal enclosure a few dozen volts above the ground the customer is standing on, draws a couple of amps, and never trips anything, because a 20 A breaker has no opinion about 2.4 A. Techs who understand a ground fault as "a short to ground" expect a trip, do not get one, and conclude nothing is wrong. The useful definition is different, and it changes what you look for.

First move when you suspect one

If a person reports a shock or a tingle from equipment, or if you feel one, that equipment is de-energized at its overcurrent device before anything else happens, and it stays off. A tingle from a grounded surface is not a nuisance; it is the leading edge of the exact mechanism described below.

If equipment is standing in water, do not enter the water and do not reach into it. De-energize from a dry location at the panel or disconnect, lock and tag under 29 CFR 1910.333(b)(2) (29 CFR 1926.417 on construction work), and prove dead using the live-dead-live sequence at NFPA 70E-2021, 120.5 before anything is touched.

Do not reset a ground-fault device more than once on a load that keeps tripping it, and never defeat or bypass one to get equipment running. A device that keeps operating is reporting a real return path through something that is not a circuit conductor, and the honest outcome is that the appliance or circuit comes out of service until the path is found.

Any live measurement described below falls under the troubleshooting exception at 29 CFR 1910.333(a)(1), which permits energized work only where the test cannot be performed de-energized, with instruments and leads rated for the system under 29 CFR 1910.334(c)(2) and an arc-flash risk assessment first under NFPA 70E-2021, 130.5.

The definition that works

A ground fault is current returning toward its source by a path that is not the circuit conductor intended to carry it.

That is it. Nothing in the definition says how much current, and nothing says the fault is violent. The magnitude is set entirely by the impedance of whatever path the current found, and that impedance can be a fraction of an ohm through a bonded metal enclosure or fifty ohms through dirt. Same fault, same voltage, currents differing by more than a hundredfold, and completely different outcomes.

What it is not

Six things get called ground faults or confused with them, and separating them out is most of the skill.

It is not the same thing as a short circuit. A short is current taking an unintended path between circuit conductors, generally at very low impedance and very high current. A ground fault is current taking an unintended path to a grounded or groundable part. The two overlap in one case - a bolted fault to a well-bonded enclosure on a solidly grounded system behaves like a short and clears like one - and the confusion comes from treating that one case as the whole category. The distinction between overload and short circuit is covered in a sibling article; this one owns the path-to-ground case.

It is not an "open ground." A missing or broken equipment grounding conductor is a defect, not a fault. Nothing is flowing anywhere it should not. What it does is remove the low-impedance path that would have made a future ground fault clear quickly, which converts a fault that would have tripped in a cycle into one that energizes an enclosure indefinitely.

It is not necessarily high current. This is the one that gets people hurt, and the arithmetic below shows it.

It is not the same as leakage. Every circuit has some capacitive leakage to ground, and equipment with filters and long cable runs has more. Ground-fault devices are set above the level normal leakage produces, which is exactly why an installation with many filtered loads on one protected circuit can accumulate enough normal leakage to cause nuisance operation without any fault existing.

It is not cleared by a ground rod. The grounding electrode's job is to reference the system to earth and give lightning and utility surges somewhere to go. Earth is a poor conductor over any distance, and the fault-clearing path is the equipment grounding and bonding path back to the source, not the dirt. A sibling article covers grounding and bonding as separate jobs; the piece that belongs here is that a fault current returning through earth is a fault that is not going to clear.

It is not always a GFCI's problem. Ground-fault protection for people and ground-fault protection of equipment operate at levels that differ by a factor in the hundreds, and an ordinary overcurrent device clears a low-impedance ground fault without any ground-fault sensing at all. Those are three different devices answering three different questions, and the table below adds a fourth.

Same fault, two return paths

A 120 V circuit protected by a 20 A breaker. The energized conductor chafes through and contacts a metal equipment enclosure. That is the fault, and it is identical in both cases below. What differs is what the current finds.

Case 1: the enclosure is bonded by an equipment grounding conductor. Total impedance around the fault loop - source, circuit conductor, the fault itself, and the grounding conductor back to the source - is low. Say the loop measures 0.3 ohms, of which the grounding conductor's share is 0.12 ohms. Fault current is 120 divided by 0.3, about 400 A. That is 20 times the breaker's rating, well into the magnetic element's range, and the breaker opens in a fraction of a cycle.

During those few milliseconds the enclosure does rise above earth by the drop across the grounding conductor: 400 A times 0.12 ohms is 48 V. That is real and it is why the speed of clearing matters, not just the fact of clearing. But it lasts a fraction of a cycle rather than indefinitely.

Case 2: there is no equipment grounding conductor and the enclosure is grounded only by a made electrode. The return path is now electrode, earth, and the service electrode. The electrical code treats a single made electrode with a resistance to earth above 25 ohms as requiring supplementation, which tells you the order of magnitude you are dealing with; assume 25 ohms at each end for this walkthrough and read the actual requirement in Article 250 of the code adopted in your jurisdiction.

Fault current is 120 divided by 50, about 2.4 A. The 20 A breaker sees 12 percent of its rating and does exactly nothing, forever. Meanwhile the enclosure sits at the drop across its own electrode: 2.4 A times 25 ohms, about 60 V above remote earth, continuously, on a piece of equipment that looks and works completely normally.

Now put a person on it, hand on the enclosure, standing on damp ground. The human body is commonly modelled at roughly 1,000 ohms hand-to-foot under wet conditions, which varies widely with contact area, moisture and path and should be used as an order-of-magnitude figure rather than a specification. That 1,000 ohm path sits in parallel with the 25 ohm electrode, giving about 24.4 ohms, so the loop becomes 49.4 ohms, total current 120 divided by 49.4, about 2.43 A, and the enclosure settles at about 59 V. The share going through the person is 59 divided by 1,000, about 59 mA.

Widely cited physiological thresholds put perception near 1 mA, loss of muscular control in the region of 10 to 16 mA, and meaningful risk of ventricular fibrillation above roughly 50 to 100 mA for a hand-to-foot path lasting a fraction of a second. Fifty-nine milliamps is inside that range, and the breaker still has no reason to move, because the total 2.43 A remains far below 20 A.

That is the entire argument for the equipment grounding conductor, stated in numbers. Case 1 and Case 2 are the same fault. One clears in under a cycle. The other is a permanent hazard that no overcurrent device in the building will ever notice.

Which device answers which question

Run Case 2 against that table. The ground-fault device would have opened at 5 mA of imbalance. The actual imbalance was 2.4 A, roughly 480 times its trip level. The device that could not see the fault at all was the only one installed.

That is also the honest scope limit on ground-fault devices: they measure imbalance between the circuit conductors, so they detect current leaving the circuit. They do not detect a person bridging the two circuit conductors, because that current goes out and comes back exactly as the device expects. A ground-fault device is not a general electrocution preventer, and describing it to a customer as one is a promise it cannot keep.

Testing without fooling yourself

Use the device's own test button. It is wired to create the imbalance internally, so it tests the sensing and the mechanism regardless of the wiring outside it.

A plug-in tester works differently: it creates a small fault from the energized conductor to the equipment grounding conductor. If that grounding conductor is missing or broken, the tester cannot make a fault, the device does not operate, and a tech reads a perfectly functional protective device as failed. That false negative is common on older installations and is itself the finding - the tester just told you the equipment grounding conductor is not there.

When a ground-fault device trips repeatedly with nothing obviously wrong, separate a real fault from accumulated leakage by unplugging loads one at a time and re-testing, rather than by replacing the device. A protected circuit carrying several filtered electronic loads can total enough normal leakage to operate a 5 mA device with no fault present anywhere, and swapping the device changes nothing.

How to verify you got this right

After any repair involving grounding or bonding, confirm the fault path exists rather than assuming it. Continuity from the enclosure back to the panel's grounding busbar is the minimum check, made de-energized under lock and tag, and it needs to be a low-resistance measurement with the meter's leads zeroed, because the numbers that matter here are hundredths of an ohm and lead resistance alone is several times that.

Then re-test any ground-fault device with its own test button and, separately, with a plug-in tester. Both passing means the device works and the grounding conductor is present. The button passing and the plug-in failing means the device works and the grounding conductor does not, which is precisely the condition that turns Case 1 into Case 2.

References

• NFPA 70 (National Electrical Code), Article 250 for grounding and bonding, the fault-current path, and grounding-electrode resistance requirements - use the edition adopted in your jurisdiction
• UL 943 for Class A ground-fault circuit interrupter trip levels
• IEC 60479 for the physiological effects of current on the human body
• 29 CFR 1910.333(a)(1) and (b)(2) - the troubleshooting exception and safe work practices for electrical work; 29 CFR 1926.417 for the construction counterpart
• NFPA 70E-2021, 120.5 (establishing an electrically safe work condition) and 130.5 (arc flash risk assessment)
• See related: The Difference Between an Overload and a Short Circuit; Grounding - The Why and the Quick Check; GFCI and AFCI Troubleshooting Reference