The Verification Venue · a rod that catches, not a magnet that summons

The Sphere That Rolls Over Your Roof

A lightning rod does not reach up and pull the storm to your house, and it does not drain the cloud. It wins the last jump of a strike that was already coming down nearby, and it paints a protected zone you can watch: roll an imaginary sphere over the roof and the terminals, and whatever the sphere cannot touch is safe. Grab the sphere below and roll it yourself.

The method has a name, the rolling-sphere method, and it is written into IEC 62305 and NFPA 780. The sphere radius is the design striking distance, set by the electrogeometric relation r = 10 · I^0.65 (r in metres, peak current I in kA). It is not the rod's "pull". Anything the rolling sphere kisses can be struck; anything it rides over is protected. Operate the bench, and one honest chain moves end to end: leader current sets the striking distance, the striking distance sets the sphere, the sphere sets the protected zone, and a weak stroke below the class minimum slips under it.

ROLL IT Roll the sphere across the roof. Where it touches, that surface can be struck.

Sphere radius r

45 m

= 10 · I^0.65, I = 10.0 kA

Min current / efficiency

91%

catches strokes I ≥ 10 kA

Ground protection radius

28.3 m

√(2rh − h²) under the tallest rod

The counter-intuitive part: the smallest sphere (LPL I, 20 m) is the strictest class. A small sphere dips into gaps a big one floats over, so it forces terminals closer together. Switch LPL IV to LPL I and watch the green shrink and the sphere sag into the ridge.

Selected rod height: 10 m

Raise the rod and a clean green wedge opens beneath it. Its attractive radius is about 3H.

Leader peak current: 30 kA

Log scale, 1 to 100 kA. Its own striking distance r_I = 10 · I^0.65 decides where it snaps.

Flash density Ng: 4.0 /km²/yr

Sets the "strikes intercepted per year" estimate from the collection area Ad.

Person / downpipe gap to conductor: 0.30 m

Feeds the side-flash panel: a correct rod can still arc sideways if this gap is under the required separation s.

Thunder is off by default and synthesized in-browser at a low, capped level. It plays only after you turn it on and only on a completed strike. Hearing-safety note: keep the volume gentle.

Attractive radius & collection area (IEC 62305-2)

tallest rod H10 m

attractive radius 3H30 m

Ad = LW + 2·3H(L+W) + π(3H)²4127 m²

strikes intercepted / yr0.017

This is the reach that does attract: any downward leader landing within about 3H prefers the rod. That is the job.

Side-flash separation (IEC 62305-3)

s = ki · kc · l / km0.10 m

ki (class) · kc · km0.04 · 0.5 · 1

length along conductor l5.0 m

your gap g0.30 m

Gap is wider than the required separation. No side-flash predicted here.

Intercepted (to a terminal)

Bypassed (into the green)

Empirical vs class

–

Roll the sphere over the bare house first (press Clear rods): it touches every corner and eave, all red, nothing protected. Raise one rod and a green wedge opens, the sphere now riding up and over. That wedge is the protected zone, and its ground-level half-width is exactly √(2rh − h²), no attraction radius anywhere in it.

Why the smaller sphere is the stricter class

Every explainer that ranks the classes gets this backwards, because it feels backwards. A bigger sphere sounds like bigger protection. It is the opposite. Because striking distance grows with current, a bigger sphere is the design distance of a stronger minimum stroke, and a big smooth sphere floats over small gaps without ever dipping in. The small LPL I sphere (20 m, the design distance of a 3 kA stroke) probes into those gaps and touches surfaces the big sphere missed, so it forces you to add terminals and space them tighter. Add a second and third rod above and drag them apart: the sphere sags between them by p = r − √(r² − (s/2)²), and the small sphere sags first. Smallest sphere, strictest spec.

"Protected" is a probability, not a promise

Fire a median 30 kA bolt and it snaps obediently to the rod. Now crank the peak current down to 2 kA and fire again, or press Show the weak-stroke bypass. A weak stroke has a tiny striking distance, r_I = 10 · 2^0.65 = 15.7 m, so its little sphere sags under the class design sphere and can scorch the roof inside the green. That is the honest residual failure: roughly 1% at LPL I up to 16% at LPL IV, the strokes weaker than the class minimum current. The interception efficiency (99, 97, 91, 84 percent) is the fraction of strokes strong enough that the design guarantees a catch. It is a probability, not a guarantee, and the storm tally converges on it as you fire.

Both popular answers are wrong, in opposite directions

The myth says a rod attracts lightning to your house. The thin correction now all over the search results over-corrects: a rod does not attract lightning at all, it only offers a path for a strike that was coming anyway. The electrogeometric truth sits between them. A tall grounded point does increase strikes to that structure, within an attractive radius of about 3H (IEC 62305-2). That is precisely its job: it converts a hit on your vulnerable roof into a hit on a terminal wired safely to ground. What it does not do is summon storms, raise the regional flash rate, or drain the cloud.

The myth-meter

False

A rod attracts storms to your town, or drains the charge out of the cloud.

This was Benjamin Franklin's own first guess in 1750, and he was wrong. A rooftop point cannot bleed a charge sitting kilometres up. Franklin himself later reframed the rod as an interceptor and conductor. The idea is resold today as Early Streamer Emission (ESE) and Dissipation Array Systems (DAS/CTS) terminals; NFPA reviewed ESE from 2000 onward and found no scientific basis, and keeps them out of NFPA 780.

True

A rod attracts the strikes already headed for this structure, within about 3H.

For the tallest rod here (H = 10 m) that reach is about 30 m. A downward leader whose final jump lands in that neighbourhood terminates on the rod instead of the roof. Attracting those strikes, and only those, is the entire point.

False

A rod raises the regional strike rate.

The number of flashes per square kilometre per year (Ng) is set by the storm, not by your roof. A rod only redistributes, within its own small neighbourhood, which point the strike lands on.

Franklin's discarded idea, and its modern zombies

"Would not these pointed rods probably draw the electrical fire silently out of a cloud before it came nigh enough to strike?"
Benjamin Franklin to Peter Collinson, 1750.

He was wrong, and honest about testing it. The rod's real value is interception, not dissipation. Yet the dissipation dream keeps coming back as ESE rods (claiming an earlier, longer streamer and therefore a larger zone) and dissipation arrays (claiming to prevent strikes). Independent reviews, including Uman and Rakov's critical survey (BAMS 2002), conclude ESE rods perform no better than plain Franklin rods, and field surveys of ESE-equipped tall buildings document repeated close-proximity bypasses. Treat the larger-zone and strike-prevention claims as marketed but unproven.

The third layer: side flash

Even a correct rod can arc sideways. If a person or a metal pipe stands too close to the down-conductor while it dumps the strike to ground, the voltage can jump the gap. IEC 62305-3 sets a required separation s = k_i · k_c · l / k_m. For a strict LPL I system with a 10 m run of a single-branch down-conductor in air, that is s = 0.08 · 0.5 · 10 / 1 = 0.40 m of clearance (a masonry path halves k_m and doubles it). Slide the gap in the bench under that separation and the side-flash panel turns red. This is why sheltering under a tall protected object can still hurt you, and why bonding, not the rod alone, is what makes a system safe.

The check: every number recomputed in front of you

Nothing here is a quoted figure. For the current class and rod the page recomputes the whole chain from the same relations the offline verifier uses:

–

Standard-audit: the four IEC radii are regenerated from their minimum currents by r = 10 · I^0.65, and the efficiencies from the first-stroke current distribution. Green check means the page's own formula reproduces the tabulated value within rounding.

class	I_min	10·I^0.65	IEC r	P(I>I_min)	IEC E

The offline gate recomputes all of it and exits 0 only if every check passes: node research/do-lightning-rods-attract-lightning/verify-do-lightning-rods-attract-lightning.mjs.

What is exactly true here, what is a model, and every free choice

The electrogeometric coefficient is a convention, and rivals differ. This page uses r = 10 · I^0.65, the relation underlying IEC 62305 that reproduces the standard radii 20 / 30 / 45 / 60 m from the minimum currents 3 / 5 / 10 / 16 kA within rounding. Rival electrogeometric models use coefficients near 6.7 to 8 and different exponents (Golde, the IEEE Std 998 form, Love, Eriksson). They give somewhat different striking distances. We name the model rather than pretend one coefficient is the single truth. The radii are fixed by the standard as 20 / 30 / 45 / 60 m by convention; the raw formula gives 20.4 / 28.5 / 44.7 / 60.6.

The efficiencies come from a distribution with real scatter, and two medians float around. The interception efficiencies (99 / 97 / 91 / 84 percent) are the tail probabilities P(I > I_min) of the first negative stroke peak current adopted in IEC 62305-1 Annex A: a log-normal with 50% value (median) 61.1 kA and log10 dispersion 0.576, equivalently a natural-log dispersion β = 1.33 (the CIGRE / Anderson-Eriksson low-current branch). That reproduces 98.8 / 97.0 / 91.4 / 84.4, matching the tabulated values within a few tenths of a percent. Note that the often-quoted "median about 30 to 31 kA" is a different figure, from Berger's directly measured tower strokes (IEEE Std 1243 / 1410); the IEC design distribution used for the efficiency table sits higher, at 61.1 kA. Both are real numbers from real datasets, and the spread between them is the honest uncertainty.

The rolling-sphere geometry is exact; the scene is a model. The protected zone is computed as the set of points no obstacle-free sphere of radius r can reach, which is the true rolling-sphere definition, and the single-rod ground protection radius that falls out, √(2rh − h²) for h ≤ r, is the textbook result. The house, the ground, and the rods are a two-dimensional side elevation. Real designs are three-dimensional and add mesh and protection-angle methods; the physics of what the sphere can and cannot touch is the same.

The leader-attachment rule is the electrogeometric model, simplified. A descending leader attaches to whichever object it first comes within its own striking distance r_I of, taken as the object with the highest attachment point. This gives the correct qualitative behaviour (strong strokes caught by the terminal, weak strokes able to slip past into a gap) but is not a full physical leader-propagation simulation. The storm tally samples currents from the IEC 62305-1 distribution and converges toward the class efficiency; the exact empirical rate depends on the geometry you build.

The collection area assumes an isolated structure. Ad = L·W + 2·(3H)(L+W) + π(3H)² is the IEC 62305-2 Clause A.2 equivalent collection area for an isolated rectangular structure with attractive radius 3H. The "strikes per year" figure is Nd = Ng · Ad · 10^-6 with an environmental/location factor taken as 1; a real risk assessment adds those factors and a full risk model.

The side-flash panel uses representative coefficients. s = k_i·k_c·l / k_m with k_i set by the protection class (0.08 / 0.06 / 0.04 / 0.04 for LPL I to IV), a representative k_c = 0.5 current-division coefficient, and k_m = 1 for air. Real values of k_c depend on the number and layout of down-conductors.

Out of scope on purpose. This page is the protection physics. It is not about how often one person can be struck (that is a Poisson survival story), and it is not about timing the flash to the thunder to judge distance. Those are their own pages.