The first time a generator bit me, it was running fine.
A rental 400 kVA set on a site outside the city, raining sideways. I leaned one hand on the canopy frame while stepping off the wet steel floor of the container — and felt it. Not a full shock. A tingle. The kind that makes you pull your hand back fast and pretend nothing happened.
Nothing was “wrong.” The set was carrying the load, voltage steady, no alarms. What was wrong was everything you can’t see from the front panel: the ground rod was driven through six inches of dry fill into broken concrete, the neutral wasn’t bonded anywhere, and the one green-yellow wire that was supposed to be the safety ground ran sixty meters back to the building’s panel and through the rebar on its way home. On paper, the machine was grounded. In the rain, I was part of the circuit.
We spent that afternoon fixing the grounding, and I never pretended nothing happened again.
Here’s the thing about generator grounding: it’s the line item everyone signs off without reading. It’s also the only thing standing between a normal Tuesday and a very bad day. And most installs get at least one piece of it wrong.
There are two separate jobs here, and people mix them up constantly.
Job one: give a fault somewhere to go, so the breaker trips. If a live wire comes loose inside the set and touches the frame, the frame has to be connected back to the source through a low-impedance path — otherwise the frame just sits there energized, waiting for someone to complete the circuit. That connection is the neutral-to-ground bond.
Job two: keep everything at the same voltage, including the dirt. Ground rods, building steel, buried pipe — that’s the earthing side. Different job. Same wire sometimes. Confusing always.
The bond is the one that scares me, because getting it wrong is so easy. The rule is simple: one bond, one place. Bond the neutral to the frame at the generator — or at the service entrance — but never both. Two bonds and your neutral current splits, part of it running forever on the ground conductors. You get nuisance tripping, warm conduits, and on one of the projects I’ve worked on, a data center that chased phantom ground faults for a week before someone drew the whole neutral path on a whiteboard and found the second bond.
Portable sets are the opposite trap. A lot of them ship with a floating neutral — no bond at all — because the manufacturer assumes you’ll plug power tools into the outlets with their own GFCI protection, where a floating system is arguably safer. Hook that same floating set up as standby power for a building without thinking, and a single ground fault won’t trip anything. The first fault sits there invisible. The second one can kill someone.
So before you argue about rods, ask the question nobody asks: where’s the neutral bonded on this install? If nobody can answer in one sentence, you don’t have a grounding system. You have a guess.
Here’s a number that ruins a lot of arguments: an eight-foot copper-clad rod driven into dry sand can read two hundred ohms. Two hundred. You might as well ground the generator to a tree.
The rod’s job is misunderstood. It is not what clears faults — the bond does that, back through the conductors, in milliseconds. The rod is for lightning, static, and giving the whole system one reference to the earth so voltages can’t float up to strange levels. A high-resistance rod still does most of that badly. A low-resistance rod does it well. What it almost never does is save you from a fault on its own.
Soil decides. The same rod that reads 25 ohms in April can read 80 in August when the water table drops. I measured one site three seasons in a row — same rod, same meter, numbers that moved like a stock chart. That’s normal. It means the “install and forget” rod your contractor put in five years ago is a decoration now.
What actually works: measure, don’t guess. A fall-of-potential test takes twenty minutes and a patient technician. If the number is bad, fix it properly — longer rods, multiple rods spaced well apart (not hammered in shoulder-to-shoulder like candles on a cake, which buys you almost nothing), ground rings, or my favorite on rocky sites: the foundation rebar. A Ufer ground — the concrete-encased electrode — beats a rod in dry ground every time, and it’s already there. You just have to connect to it before the concrete is poured, which means thinking about grounding on day one instead of day thirty.
I’ve been doing this twelve years. The mistakes don’t change. They just change sites.
Mistake one: the rod nobody tested. Driven in, wire attached, box ticked, never measured. Half the rods I test read over 100 ohms. The fix is a meter, not a bigger rod.
Mistake two: the neutral bonded twice — or never. Portable set feeding a building, floating neutral, no bond anywhere. Or a standby set bonded at the generator and again at the main panel, with neutral current living permanently on the ground wire. One bond, one place, drawn on paper.
Mistake three: the ground wire that isn’t there anymore. Ground conductors get borrowed for other jobs, cut during maintenance, “temporarily” removed and never put back. On remote sites they get stolen for the copper — I’ve seen that twice. A grounding system is a chain. Every maintenance visit should end with somebody confirming the chain is still connected.
That’s the whole list. It’s not a long list. That’s what makes it embarrassing.
People think grounding is about the machine. Half of it is about you — where you stand.
Touch voltage is the difference between the energized frame and the ground under your boots. In a dry generator room, you might never notice a problem. In the rain, on wet concrete, with a high-resistance rod and a loose bond, that same fault puts real voltage across your body. The generator doesn’t care. Physics doesn’t negotiate.
Step voltage is the sneaky cousin: voltage across the ground itself near a fault or a lightning strike, one foot to the next. You don’t touch anything and it still gets you. This is why substations and big genset yards use ground grids and gravel — to flatten the voltage gradient. For a containerized set on a wet site, the practical version is simpler: bond everything metal, one common ground reference, no floating steel.
And that brings me to the building’s steel. Everyone wants to use it as the ground — it’s big, it’s buried, it’s free. Sometimes that’s right. But steel connected at two ends to two different ground systems becomes a conductor for stray current, and now the building itself is carrying neutral current it was never meant to carry. I’ve measured it. So connect to the steel — once, at one point, by design — or don’t connect to it at all. Half a ground is worse than none, because at least “none” is honest about what it is.
One more thing people skip: surge protection at the transfer switch. The ground rod handles the lightning’s ego, not its details. A strike a street away can still push enough through to cook control boards. A good automatic transfer switch installation includes surge protection on both sides. The grounding system and the surge devices are a team. Solo, neither one finishes the job.
So where does that leave you? Honestly, it’s a short checklist, and you could do worse than taping it inside the generator room door.
Know where the neutral is bonded — one place, written down. Test the ground rod in the dry season, not just the wet one. Confirm the ground wire is still connected after every service visit. Put GFCI/RCD protection on the portable sets and ground-fault protection where the three-phase world requires it. And if you’re sizing a new set, think about grounding on day one, when the foundation rebar is still exposed and the fix costs nothing.
Grounding is the cheapest part of the install and the hardest part to fix later. The tingle I got in the rain was the machine’s way of telling me that. Listen to it before it has to say it louder.
If you’re planning an install and want a second pair of eyes on the grounding design before the concrete goes in, talk to us. Cheap insurance.
Email：[email protected]
Work：Monday to Friday,9:30-18:30
