The IEC 62305-3 separation distance, worked through

1. What the separation distance is protecting against

When lightning current runs down a conductor, that conductor's potential rises far above everything around it. Metalwork inside the building — a pipe, a cable tray, a duct, structural steel — sits near earth potential. If the two are close enough, the gap flashes over, and the lightning current takes an unintended route through the inside of the building.

That is the failure the separation distance prevents. It is a required clearance: the minimum gap at which the insulation between the two can be relied upon not to break down. Get it wrong and the external LPS becomes a delivery mechanism rather than a diversion.

2. The three coefficients

The calculation scales a conductor length by three factors. Each answers a different question, and knowing which one you can actually influence on a given project is most of the skill.

• k_i — how much current you are designing for. It follows from the LPS class, which follows from the risk assessment. Class I protects against the most severe strikes and carries the largest coefficient; Class IV the smallest. You do not choose this freely — it is an output of the preceding step.
• k_c — how the current divides. The fraction of the total that flows in the particular conductor you are standing next to. More down conductors means each carries less, so k_c falls. This is the coefficient you can design against, and section 4 is about it.
• k_m — what is in the gap. Air is the reference case, and it is the best case. This coefficient sits in the denominator, and every material other than air lowers it — so a gap filled with concrete, brickwork or timber doubles the clearance the standard requires, and a purpose-made insulating stand-off still requires more than air alone. This catches people out, because "insulating material" sounds like a credit and is in fact a penalty: a solid dielectric bridging the gap flashes over more readily than the same distance of open air, so the distance has to grow to compensate. Filling a tight gap is not a way out of it.
• l — the length that matters. Measured along the conductor from the point you are considering to the nearest equipotential bonding point or the earth termination. This is the input most often got wrong, because it is a path length along the conductor, not the straight-line distance between two things.

One useful consequence of that last point: bonding higher up the structure shortens l for everything above it. On a tall building with a ring conductor at an intermediate level, the separation requirement above that level is computed from a much shorter length than it would be if the only bonding were at the ground.

3. A worked example

This is a published design from a well-known guide to the standard, which we ran through our own engine as a validation case. Using someone else's worked example rather than inventing one is the point: the figures can be checked against a source that is not us.

The building is 32 × 32 m in plan and 15 m tall — a 128 m roof perimeter — at LPS Class IV, with concrete walls, a 2 m protective-angle air rod and Type A earthing in 160 Ω·m soil. Working down the chain, Class IV gives the smallest of the k_i values; concrete gives the solid-material k_m, which is the doubling described above — the same building with an air gap would need half this clearance; the down-conductor spacing limit for Class IV is 20 m, so the 128 m perimeter needs seven conductors.

The published design takes k_c = 1 — its maximum — because it invokes the dissimilar-resistance case described in section 6. Holding that assumption fixed, the arithmetic reconciles exactly:

The practical reading of 1.2 m: any pipe, tray or duct running within 1.2 m of a down conductor at that height must be bonded to it, or moved. On a real façade that is a large clearance, and it is why this calculation belongs at the start of coordination rather than at the end.

Left to design the same building itself, our engine does not assume k_c = 1 — the dissimilar-resistance case is a condition to be established on site, not a default. It reads the ordinary current-division value for the seven down conductors it fits, and returns a smaller distance. This is that design, drawn by the tool:

Both figures are defensible and the gap between them is the point: the assumption, not the arithmetic, is what moves this number — by better than a factor of two on one building. A tool that silently picked either one for you would be hiding the only decision that mattered. Ours computes the even-division case and prints the k_c = 1 caveat as a note, so the engineer resolves it rather than inherits it.

One further difference worth stating: our engine fits seven down conductors where the published example uses six — the source rounds 128 / 20 down, which puts the spacing over the Class IV limit, and we round up instead. Both differences are recorded in full on our validation page, including the reasoning, because a validation record that only lists the matches is not a validation record.

4. k_c, and the two ways to compute it

The simplified approach takes a single current-division value for the whole conductor length, read from how many down conductors the structure has. It is quick, it is conservative, and it is what a one-page calculator gives you.

The general approach traces the current through the actual geometry. Where down conductors are interconnected by ring conductors at each level, the current entering at the top divides again at every ring it passes, so the fraction in any one segment falls as you descend. The calculation becomes a sum over segments — each with its own length and its own division factor — rather than one multiplication.

The difference is not academic. A published worked example on a 60 × 60 × 7 m Class II building with 24 down conductors traces the path to a roof-mounted air-conditioning unit across five segments; the segment sum comes to 7.216, and the resulting separation distance is 0.87 m. Our engine implements this same summation and reproduces that figure. The simplified route on the same building would have demanded considerably more clearance for a unit that is, in reality, well protected by the mesh around it.

This is what a consultant does by hand on a difficult building, and it is the calculation that turns "that plant platform cannot go there" into a design that gets built. It needs the conductor length and the current-division geometry as inputs — which is precisely why the free single-form calculators cannot produce it. Their input sets never ask.

5. The 2024 change that will make your old example disagree

If you check a current tool against a handbook written for the previous edition — and you should — one case will not match, and it is worth knowing which before you conclude the tool is broken.

The current-division value for exactly two down conductors. The 2024 standard's Table 13 gives a higher figure for that case than the design guides written against the previous edition print, so the same building with two down conductors comes out needing a larger separation distance here than an older book will tell you. The values for one down conductor and for three-or-more agree with those guides, which is why a clean cross-check should use an example with one or with three-or-more.

Worth being blunt about the direction: computing to the 2024 table makes the requirement stricter in that case. A two-down-conductor design carried over from an older calculation may not meet it, and nothing will flag that. Separately, the minimum earth-electrode length curve for Class II also disagrees with the older guides — so if you are validating earthing against an older source, use Class I, III or IV.

Everything else in this chain reconciled against those guides exactly, and should match a good older source too: the rolling-sphere radii and mesh sizes, the down-conductor spacings, k_i, k_m, the separation distance for one or three-or-more down conductors, the material cross-sections and the Type A/B electrode geometry. A mismatch in any of those is a real finding, and we would want to hear about it.

6. When the distance cannot be achieved

Often it cannot, particularly on retrofits and on dense plant rooms. There are three compliant answers, and it is worth being clear that "close enough" is not among them.

• Bond it. If the metalwork cannot be far enough away, connect it deliberately so there is no gap to flash across. This has consequences — bonded metalwork now carries part of the lightning current, which is a question for the internal protection design rather than something to wave through.
• Increase the distance. Either physically, or by reducing what is required: add down conductors so the current divides further, bond at an intermediate level so l shrinks, or compute k_c by the general segment-by-segment route rather than the simplified one. Note what is not on this list: filling the gap with insulating material increases the requirement, as section 2 explains.
• Use an electrically insulated LPS. Where the distance genuinely cannot be satisfied, the standard's own alternative is an insulated down conductor with an equivalent separation distance declared by the cable manufacturer, which the design then works to. This is a specified product with a tested figure behind it — not the same thing as putting insulation in a gap and assuming it helps.

Two conditions worth checking before you rely on a number. With Type A earthing the current-division values assume neighbouring electrode resistances are within a factor of two of each other; where they differ by more, the worst case applies and the required distance rises sharply. And at high elevation an altitude correction applies to the withstand, which is easy to forget on mountain and plateau sites.