Grounding and Bonding as Two Different Jobs
Why this matters
Almost every tech can say that grounding and bonding are different. Far fewer can say what breaks when you use one to do the other's job, and that gap produces a specific, dangerous repair: a shock complaint answered with a ground rod. It feels like the right answer, it sometimes even reduces the symptom for a few weeks, and it cannot clear a fault, because earth is not a fault-return path. Current returns to its source, not to the planet.
That single sentence is the whole distinction, and the arithmetic below is what makes it undeniable.
A tingle is a live-chassis call: what to do before you diagnose
Somebody feeling a shock from equipment metal means that metal is at a different potential from what they are standing on. Treat it as energized until proven otherwise.
Do not touch it again to see how bad it is, and do not let anyone demonstrate it for you. Keep people, children and animals away from the equipment. If the location is wet ground, a pool, a spa, a dock or a wet slab, keep everyone out of and off the water and the wet area entirely before any measurement, because contact current through water can immobilise a person before they can move themselves out of it.
De-energize the branch at the panel, lock and tag it under 29 CFR 1910.333(b)(2) - 29 CFR 1926.417 where the work falls under construction - and prove dead live-dead-live per NFPA 70E-2021, 120.5 before you work on it. Any measurement you take while it is still live, including a touch-voltage reading, is energized work under 29 CFR 1910.333(a)(1) and needs a meter and leads rated CAT III or above at the voltage present.
Keeping the customer clear is not an OSHA duty, since those duties run to your own employees. It is the ordinary duty of care, and it matters just as much.
The call, and the fix that could not have worked
A well pump control cabinet, outdoors, on a 120 V branch circuit. The homeowner feels a tingle touching the cabinet while standing on wet ground. A previous visit answered it by driving a supplemental ground rod at the equipment and bonding the cabinet to it. The complaint eased and then came back.
Two things are worth separating here, because the previous tech was not being lazy. Driving that rod did change the numbers slightly, by giving the cabinet a nearer connection to local earth than it had. And it evaporated with the weather, because soil resistivity swings by roughly an order of magnitude with moisture and temperature. A rod that measures well in a wet spring measures far worse in a dry August, so any improvement bought from earth contact is a seasonal number, not a repair.
If you drive a rod on any job: call the underground locate service first. Driving an eight-foot rod into an unlocated buried gas or electric service is a fatality mechanism, not a code violation.
The arithmetic that kills the ground-rod theory
Measure the rod. Say it reads 25 ohms to earth, which is a respectable value; NEC Article 250 requires a single made electrode to be supplemented unless it tests at 25 ohms or less, so this rod is at the boundary the code treats as acceptable for a single electrode.
Now ask what that rod does with a fault. An ungrounded conductor touching the cabinet, with the rod as the only return path, drives 120 / 25 = 4.8 A through the earth.
A 20 A breaker will carry 4.8 A indefinitely. Not slowly, not reluctantly: forever, with no indication of any kind. The cabinet stays energized at close to line potential relative to any earth further away than the immediate soil around that rod, and the only thing between the customer and that potential is the resistance of their shoes and the ground they are standing on.
Compare a properly bonded path. An equipment grounding conductor run with the circuit conductors back to the point where the system is bonded at the source gives a ground-fault loop impedance that is a fraction of an ohm; say 0.35 ohms measured with a loop tester on this circuit after repair. The same fault now drives 120 / 0.35 = 343 A, which is 17 times the breaker's 20 A rating and lands squarely in its instantaneous magnetic region. The breaker opens in a fraction of a cycle and the cabinet is dead before anyone can walk over to it.
Those two paths differ by 343 / 4.8 = 71 times in current, and that ratio is the entire argument. The bonded path is not a better ground. It is a different mechanism doing a different job.
Scale it against the human numbers to see how far outside the safe range 4.8 A is. IEC 60479-1 describes the effects of current through the body by magnitude and duration: perception starts near 1 mA, the let-go range is commonly cited near 10 mA, and the risk of ventricular fibrillation rises steeply above roughly 30 mA sustained. A Class A ground-fault circuit interrupter under UL 943 is built to trip at about 6 mA for exactly this reason. The 4.8 A flowing through that earth path is 800 times the current a GFCI is designed to interrupt, and the overcurrent device is the only thing watching it.
What was actually wrong, and why nothing indicated it
The circuit was an older two-wire feed relying on a metal raceway as its equipment grounding path, and a buried coupling in that raceway had corroded through. The cabinet had no low-impedance path back to the source at all.
Nothing indicated it because nothing tests for it in normal operation. The pump ran. The breaker held. A receptacle tester at the panel end reported correct wiring on the circuits it could reach. An equipment grounding conductor carries no current at all until the day it has to carry all of it, which means a broken one is invisible for its entire life up to the moment it is needed. That is the property that makes it worth deliberately testing rather than assuming.
The two failures compounded. Had the raceway path been intact, the fault would have cleared the instant it developed and there would have been no tingle to report. Had the previous visit tested the return path instead of adding an electrode, it would have been found on that visit.
Confirming the repair without creating a fault
You do not prove a fault path by making a fault. You prove it with instruments:
- End-to-end continuity of the new equipment grounding conductor, measured de-energized, from the equipment enclosure to the bonded point at the source. Use a four-wire low-resistance method for this, because the value you care about is a fraction of an ohm and a handheld meter's own leads are the same order of magnitude as the reading.
- Ground-fault loop impedance with a loop tester, energized, at the equipment. This one measurement is the answer to the question the whole article asks: it reports the impedance of the actual path a fault would take, including every connection along it. It read 0.35 ohms here.
- Touch voltage on the cabinet relative to a reference point away from the equipment, with the pump running. It should be a fraction of a volt. Before the repair it was near line potential.
- GFCI operation, if one is present, using a tester that injects a defined fault current rather than only the device's own test button. The button proves the mechanism; the injected test proves the mechanism plus the path.
Record all four. A ticket that says "grounded properly" is worth nothing to the next tech; four numbers with their units are worth the whole visit.
The two jobs, side by side
| Grounding (connection to earth) | Bonding (metal to metal, back to the source) | |
|---|---|---|
| What it is for | Limiting the voltage imposed on the system by lightning, surges and accidental contact with higher-voltage lines, and stabilizing system voltage to earth in normal operation | Providing a low-impedance path for fault current back to its source so the overcurrent device operates, and holding touchable metal at a common potential |
| Path material | Soil | Metal: conductors, raceway, enclosures, bonding jumpers |
| Typical impedance | Ohms to tens of ohms, and seasonal | A fraction of an ohm, and stable |
| Fault current it carries | Negligible | All of it |
| What happens if it is missing | Surge and lightning protection degrades; nothing obvious in daily operation | Faults do not clear; enclosures can sit energized indefinitely |
| How the failure announces itself | It does not | It does not |
The bottom row is why both get tested rather than inspected. Neither system gives any indication in normal service, so their condition is only ever as good as the last time somebody measured it.
Where bonding has a second job: equipotential
Clearing a fault fast is enough to protect a person standing on a dry floor. It is not enough where a person is in water or in continuous contact with conductive surfaces, because the current needed to hurt them is thousands of times smaller than the current needed to trip a breaker, and it can flow through them without ever being a fault the breaker can see.
That is why pools, spas and similar installations require an equipotential bonding grid: the objective is not to clear anything, it is to eliminate the difference in potential between everything a person can touch at once, so there is no voltage available to push current through them. A sibling article covers the specific requirements. The reason it exists belongs here: bonding does two separate jobs, and equipotential bonding is the one that protects people from voltages far below what any overcurrent device will ever respond to.
References
- NEC Article 250 for grounding and bonding requirements, including electrode resistance and equipment grounding conductors
- IEC 60479-1 for the effects of current on the human body by magnitude and duration
- UL 943 for Class A ground-fault circuit interrupter trip characteristics
- 29 CFR 1910.333(a)(1) for energized-work justification and 29 CFR 1910.333(b)(2), with 29 CFR 1926.417 for construction, for de-energizing, lockout and tagging
- See related: Why Neutral and Ground Are Not the Same; Pool and Spa Electrical Bonding Reference; How to Use Voltage Drop to Find a Bad Connection