The Resistor Nobody Specs — Until a Ground Fault Burns the Stator

A stator doesn’t die quietly. When a ground fault burns one, the smell is unmistakable — hot iron and cooked varnish — and by the time you smell it, the damage is done.

Tuesday morning, nine years ago, a paper mill called. Their 1 MW set was down. The night shift had smelled ozone, then smoke, then nothing — the generator had tripped on differential and never came back. We pulled the end covers and found a crater in the stator iron where a winding had shorted to the core. The rewind quote ran into five figures. The repair took eleven weeks.

The mill’s system was ungrounded. The first ground fault had happened months earlier and nobody knew. When the second fault landed on another phase, it completed a phase-to-phase path straight through the stator. Two invisible faults, one burned machine.

A resistor the size of a shoebox, wired between the generator neutral and ground, would have changed the entire story. That resistor has a name: a neutral grounding resistor, NGR. And in my experience, it is the most skipped line item in generator specifications — until a site learns what it costs to skip.

Solid, High-Resistance, or Floating — Pick Your Poison

Every generator neutral gets grounded one of three ways. Each choice trades one problem for another. There is no free option. Here is the honest version.

Solidly grounded means the neutral is bolted straight to ground. A ground fault becomes a dead short — thousands of amps, breakers trip in milliseconds. That sounds good. It is good, right up until the arc melts stator iron. Solid grounding clears faults fast and lets the fault do maximum damage in the milliseconds before the breaker opens. On small low-voltage sets, that trade is usually fine. The machine is cheap, the breaker is fast, and the arc doesn’t have time to eat much iron.

High-resistance grounding puts an NGR between neutral and ground and limits fault current to a trickle — typically five to ten amps on low-voltage systems. The fault doesn’t trip anything. It alarms. A relay — usually labeled 59G, watching neutral-to-ground voltage — tells you a fault exists and lets you shut down in an orderly way. The arc energy drops to almost nothing. Stator iron survives. And the building stays on power until you decide otherwise, which for a standby set is the whole point of having it.

Ungrounded means exactly that: the neutral floats. The first ground fault draws only the system’s capacitive charging current — milliamps — and nothing trips. Sounds elegant. It is a trap. The fault sits there, invisible, degrading insulation and waiting for a second fault on another phase. When the second one arrives, you get a phase-to-phase fault with no current limit at all. That is what burned the paper mill’s stator. Ungrounded is a bet that you will find the first fault before the second one finds you. I have watched that bet lose too many times to take it.

Resistor grids like these set your let-through current. Five to ten amps on low-voltage systems is the usual neighborhood.

So Which One? My Take

For a standby generator that exists to keep a building alive, high-resistance grounding wins more often than not. A solid ground fault on a solidly grounded 480-volt system can dump enough energy to damage the stator before the breaker clears — and if the fault is on a feeder downstream, the whole set trips and the building goes dark. With an NGR, the fault current is too small to hurt anything, the alarm tells you where to look, and you transfer load or shut down on your terms.

That said, solid grounding is not wrong. On smaller sets — say under a few hundred kVA — the fault energy is manageable, protection is simple, and ground-fault tripping is predictable. It is the honest choice when the machine is cheap enough that fast tripping beats fault survival. (The grounding electrode and bonding side of this is covered in our generator grounding safety guide.)

An old-timer I worked with for years disagreed with me on this. He said a generator that doesn’t trip on a ground fault is a generator lying to its operator. He had a point — alarming without tripping only works if somebody actually responds to the alarm. An alarm nobody hears is just decoration. But I would rather have the choice than have it made for me by a breaker at 2 a.m.

Sizing the Resistor Is Simpler Than It Looks

An NGR is sized by two numbers: the current it lets through and how long it can stand the heat. The current is usually picked to be slightly larger than the system’s capacitive charging current — on low-voltage systems, five amps covers most installations, ten for larger ones. The point is to keep fault current above the noise floor so the relay sees it, but low enough that nothing gets hurt.

The time rating matters more than people think. A ten-second rating is standard — the resistor only has to survive until you find the fault and shut down. But if your site’s procedure is “alarm, investigate, keep running while we look,” you need a continuous rating, which means a bigger, more expensive resistor. Match the rating to your actual response procedure, not the brochure’s.

One more piece that gets skipped: the resistor monitor. An NGRM — neutral grounding resistor monitor — watches the resistor’s continuity. Resistors fail open. When one does, your carefully designed high-resistance-grounded system silently becomes an ungrounded one, and you are back to betting on finding the first fault. I have walked onto sites where the resistor had been open for years. Nobody knew. The monitor costs a fraction of the resistor. Spec it.

The single-point ground connection lives in the generator terminal box. If the neutral is grounded here and again at the service entrance, you have a loop — not a ground.

The Mistake I Find on Half the Sites I Walk

The single-point rule: a generator neutral gets grounded at one point. One. The most common violation I see is the neutral grounded at the generator and again at the service entrance or the transfer switch. Two ground points make a loop, and loops carry circulating current — nuisance trips, relay confusion, and in paralleled sets, currents that wander between machines. (Our parallel operation guide gets into what circulating currents do to paralleled sets.)

Three more, while I am listing sins:

• The NGR bypassed with a jumper “for testing” that never came off. Three years later it is still there, and the system is solidly grounded with nobody the wiser.
• No resistor monitor. Covered above. Resistor fails open, system floats, first fault goes invisible.
• The alarm wired to a panel nobody looks at. Ground-fault indication that ends in an unmonitored SCADA point is the same as no indication.

NFPA 110, the standard that governs emergency power, wants ground-fault indication on the generator — not necessarily a trip. That is the code quietly agreeing with the NGR approach: know about the fault, then decide. (The code side of emergency power systems is covered in our NEC 700/701/702 guide.)

What I’d Actually Spec

If someone asked me to spec grounding for a new standby installation tomorrow: high-resistance grounding with an NGR sized at five to ten amps, a continuous time rating if the site runs manned, ten-second if it does not, a resistor monitor, the neutral grounded at the generator and nowhere else, and a ground-fault alarm that goes somewhere a human will actually see it.

And the insulation testing still matters — a ground fault starts as weak insulation, and weak insulation shows up on a megger long before it shows up as an alarm. (See our alternator insulation testing guide for the procedure.)

The paper mill eventually added NGRs to both of its sets. The resistor cabinets cost less than a week of the rewind’s downtime. Eleven weeks without a megawatt teaches you arithmetic fast. If your generator neutral is floating — or grounded in two places, or “protected” by an alarm nobody monitors — go look at it this week. The resistor is the cheapest insurance in the whole installation. Talk to us if you want help sizing one for your set.