The men who get hurt worst in an arc flash usually never touch anything. That's the part nobody tells you in the first year. You picture electrocution — current through the chest, the classic story. But the injuries that fill burn units come from a different mechanism entirely: a man standing in front of an open panel, screwdriver two inches from a bus bar, and something goes wrong in the air in front of him. He is a bystander to the event that disfigures him. He never completed a circuit. He was just standing too close.

That distance — the one that separates "I saw a flash" from "they airlifted me" — has a name in the Code. It's called the arc flash boundary, and it is not a safety-culture abstraction. It is a calculated radius, in inches or feet, around a specific piece of energized equipment, and it is derived from physics you can follow start to finish. Most electricians have heard the phrase. Fewer could tell you what number it's solving for, or why the boundary around a 480V motor control center can be larger than the one around a 13.8kV switchgear cabinet.

The thing that burns you isn't electricity

Start with what an arcing fault actually is. A short circuit through metal is a conducted event — current flows through copper, the breaker sees it, the breaker opens. An arcing fault is different. Something initiates a path through air: a dropped wrench, a loose strand, a rodent, condensation on a dirty insulator. Air is a lousy conductor until it isn't. Once ionized, it becomes plasma, and plasma conducts.

What happens next is a sustained arc drawing fault current through a channel of superheated ionized gas. The temperature at the arc terminals is routinely cited in arc flash literature as reaching tens of thousands of degrees — figures around 35,000°F get quoted, several times the temperature of the sun's surface. And here's the mechanism that matters: the copper conductors at the arc terminals don't melt so much as they vaporize, and copper going from solid to vapor expands by a factor commonly given as roughly 67,000 to 1.

So you have three things leaving that cabinet at once. A blast of thermal radiation. A pressure wave from the volumetric expansion. And a spray of molten metal droplets moving with that wave. None of them require you to touch a conductor. All of them are governed by distance.

Incident energy: the number the boundary is built on

The quantity that anchors the whole system is incident energy — the thermal energy delivered to a surface at a given distance from the arc, expressed in calories per square centimeter (cal/cm²). It is energy per unit area, which is exactly what a burn is.

And the threshold that anchors that is 1.2 cal/cm². That is the value taken as the onset of a second-degree burn on bare skin — the point where the tissue blisters rather than just reddens. NFPA 70E and IEEE 1584 both build outward from that number.

So now the arc flash boundary has a definition that isn't hand-waving: it is the distance from the potential arc source at which the incident energy drops to 1.2 cal/cm². Inside that radius, an arc flash at that equipment would deliver enough thermal energy to give an unprotected person second-degree burns. Outside it, it wouldn't. That's the whole idea. The boundary is not a comfort zone. It's a burn line.

Which means it isn't a fixed number, and this is where intuition fails people. The boundary is a function of two variables that have nothing to do with how scary the voltage sounds:

How much fault current is available. A panel fed by a big transformer with a short, fat feeder can deliver enormous current into an arc. The same panel at the end of a long run, or fed from a small transformer, can't. The arc's power is roughly arc voltage times arc current — more available current, more power dumped into the plasma.

How long the arc burns. Incident energy is power multiplied by time. This is the variable that surprises people, and it's the one that makes the numbers behave counterintuitively. The clearing time of the upstream overcurrent device is the multiplier on everything. A device that clears in three cycles — one twentieth of a second — produces a fraction of the incident energy of one that takes half a second to decide, all else equal.

And there's a third factor that's really an assumption baked into every calculation: working distance. Incident energy falls off sharply with distance — not linearly, but steeply, on the order of an inverse-square-ish relationship modified by the geometry of the enclosure. The labeled incident energy on a piece of equipment is stated at a specific working distance, typically 18 inches for low-voltage panelboards, roughly where a person's chest and face sit when they're working on it. That number is not what your hands see. Your hands are much closer, and the energy there is dramatically higher.

Why the 480V panel can be worse than the 13kV cabinet

Here is the fact that reorganizes how you think about this. Take two pieces of equipment. One is a 480V motor control center on the load side of a big service transformer, with heavy available fault current and an upstream breaker that takes a leisurely amount of time to clear. The other is medium-voltage switchgear with fast relaying that kills an arc in a few cycles.

The 480V bucket can carry the larger incident energy and the larger arc flash boundary. Not because 480 is more dangerous than 13,800 — it obviously isn't, as a shock hazard. But incident energy doesn't care about your intuition about voltage. It cares about current times time. Medium-voltage gear tends to get fast, purpose-built protective relaying because engineers took it seriously. Low-voltage distribution gear frequently sits under a molded-case breaker whose long-time or short-time characteristic lets an arcing fault — which may draw less current than a bolted fault, and therefore land lower on the trip curve — burn for a surprisingly long stretch before anything opens.

That's the trap. An arcing fault often doesn't look like a dead short to the breaker. Arc impedance limits the current. The breaker sees a fault that's high, but not high enough to slam into the instantaneous region, so it sits in the time-delay band and waits. Meanwhile the arc has all the time it needs.

This is why arc flash mitigation is so often a protection problem rather than a PPE problem. Reducing clearing time — maintenance switches that temporarily defeat time delay, current-limiting fuses, arc-energy-reducing settings — attacks the term that scales linearly with energy. Adding a layer of arc-rated clothing only changes what happens after the physics has already run.

What the boundary asks of you

The arc flash boundary is a distance. The PPE is a separate question, keyed to the incident energy at the working distance, and rated in the same cal/cm² units — an arc rating on a garment says how much incident energy it can absorb before the wearer crosses that second-degree threshold. Arc-rated is not the same as flame-resistant. Ordinary synthetic clothing under an arc-rated shirt can melt to skin. This is the detail that gets lost.

And the honest conclusion NFPA 70E keeps arriving at: the reliable protection is de-energizing. Everything else — labels, boundaries, cal ratings, face shields — is what you do when a genuine, documented reason exists to work on something live. The boundary tells you where the burn starts. It does not make it a good place to stand.

Your next moves

  • Read the label before you open the door, not after. Every piece of equipment covered by NEC 110.16 arc flash marking requirements should carry one. Find the incident energy value, the working distance it's stated at, and the arc flash boundary distance. If it says 8 cal/cm² at 18 inches, that number is for your chest — not your hands.
  • Ask when the arc flash study was last updated. Incident energy depends on available fault current and upstream protective device settings. If the utility upgraded the transformer, or someone changed a breaker setting, or a generator was added, the label on the door may be describing equipment that no longer exists.
  • Find out whether your upstream device has an arc-energy-reduction setting — a maintenance mode, an instantaneous override, a zone-selective interlock. If it does and nobody uses it, you're carrying risk that a switch flip would remove. Clearing time is the term you can actually shrink.
  • Audit what's under your arc-rated shirt. Polyester undershirts, synthetic socks, nylon-blend base layers. Arc rating on the outer layer doesn't stop meltable fabric underneath from doing what meltable fabric does.
  • Before your next live measurement, physically pace the arc flash boundary distance on the floor and notice who else is inside it. Boundaries apply to the apprentice holding the flashlight, too. They usually have no PPE at all.

Most of this comes down to numbers you need at the panel, not back at the truck — clearing time, available fault current, conductor ampacity, voltage drop on the feeder that's making your fault current lower than the study assumed. Voltly was built for exactly that moment: the NEC reference, the ampacity tables, the voltage drop and conduit and box fill math, in your hand, offline, in a basement with no signal. It won't run your arc flash study — that's an engineer's job with real fault-current data. But it will keep every other calculation you need from becoming a guess, and guesses are how bad days start. Have a look at Voltly if you'd rather stop borrowing someone else's dog-eared tables.