Cable Sizing to BS 7671: Which Check Actually Decides the Cable

A cable is not sized by BS 7671. It is sized by one of the checks BS 7671 requires — whichever one asks for the most copper — and every other check simply comes along for the ride. Knowing which one it was is the most useful single fact about a circuit, because it tells you what to change. A run held up by voltage drop does not improve with a different protective device. One held up by grouping does not improve with a shorter route.

The short version: work out the smallest conductor that satisfies each check independently, and the cable is the largest of those answers. That check governs; the gap to the next one is your margin. Two of the constraints — cumulative voltage drop and fault level — belong to the path rather than to the circuit, which is precisely why a row-by-row calculation misses them.

A table of the seven checks a BS 7671 cable calculation makes, one per row, against the ladder of available conductor sizes. Each row carries a bar that begins at the smallest conductor satisfying that check and runs to the right, so a longer bar means an easier check. For a 10 amp lighting circuit 120 metres long, overload coordination, current-carrying capacity, the CPC withstand and the minimum size are all satisfied at 1 square millimetre, and earth fault loop impedance at 2.5. Voltage drop is not satisfied until 10 square millimetres, three sizes further up the ladder, so voltage drop is the check that decides the cable. — The cable is decided by whichever check asks for most. Here it is voltage drop, three sizes above what everything else would accept.

How Many Checks Are There, Really?

Ask around and you will be told five. It is worth knowing that BS 7671 publishes no numbered list of checks at all — "the five checks" is trade shorthand, not a citation, and this article used to repeat it. Counting what a compliant calculation actually has to demonstrate gives seven, and two of them are partly the same inequality.

A list of the seven checks the calculator emits, each with its regulation number and what it proves: overload coordination under 433.1.1, current-carrying capacity under 523.1, voltage drop under 525.1, earth fault loop impedance under 411.4.204, the protective conductor withstand under 543.1.3, the line conductor withstand under 434.5.2, and the minimum conductor size under 524.1. A bracket joins the first two, because the right-hand half of Ib less than or equal to In less than or equal to Iz is the same inequality as the capacity check. — The familiar "five checks" is trade shorthand. The count that matters is six distinct requirements, and Reg 434.5.2 is the one usually missing.

The overlap is easy to miss. Reg 433.1.1 requires I_b ≤ I_n ≤ I_z; the right-hand half of that, I_n ≤ I_z, is the current-carrying capacity check of Reg 523.1. Written as two rows on a schedule they look like two hurdles; they are one requirement counted twice, plus a separate statement about the device (I_b ≤ I_n) that no conductor size will ever fix.

The genuinely separate ones are voltage drop (525.1), earth fault loop impedance (411.4.204), the minimum conductor size (524.1, Table 52.3), and two thermal withstand checks — not one. Reg 543.1.3 is the adiabatic for the protective conductor; Reg 434.5.2 is the same relationship applied to the line conductors, with a different k. They are frequently collapsed into a single "adiabatic check" on the CPC, which quietly drops a requirement.

A Correction Worth Making: the Method Is Not a Factor

One thing that often gets written down wrongly, including in an earlier version of this page: the installation method is not a derating factor. The reference method — Method C clipped direct, Method E in free air, Method B in conduit on a wall — selects which column of the Appendix 4 table you read. The correction factors C_a, C_g, C_i and C_f are multipliers applied to whatever that column gives you. Treating the method as a factor invites applying it on top of a column that already assumes it, and makes a change of method look like a scaling when it is a replacement.

Which Check Governs Is a Property of the Job

There is no general answer to "which check decides cable sizes". Four entirely ordinary circuits, run through the same engine, come out decided by four different constraints:

Four cards, each an ordinary circuit sized by the same engine, each decided by a different check. A 10 amp lighting circuit 120 metres long is governed by voltage drop. A 63 amp XLPE armoured sub-main 90 metres long is governed by earth fault loop impedance. A 32 amp circuit only 12 metres long but at 45 degrees ambient and bunched with five others is governed by overload coordination and current-carrying capacity together. A 25 amp circuit 10 metres long behind an MCCB on a board with a 16 kiloamp fault level is governed by the line conductor short-circuit withstand of Regulation 434.5.2. — Same standard, same engine, four different binding constraints. Which one binds tells you what to change.

The margin matters as much as the identity. On the 120 m lighting circuit, voltage drop asks for 10 mm² where everything else is satisfied at 2.5 mm² — three sizes up the ladder, and the entire cost difference on that run is being spent on one check. That is not a criticism of the design; it is the thing to know before someone proposes rerouting to save 10 metres, or argues about the device curve.

It is also why "size it and move on" is a poor habit on a job with any repetition. If forty circuits are all governed by grouping, the containment layout is the design decision, and no amount of care circuit-by-circuit will find that.

Voltage Drop Belongs to the Path, Not to the Cable

This is the one that catches people, and it is not an arithmetic problem. Reg 525.1 sets the limit on the drop between the origin of the installation and the equipment — 3 % for lighting, 5 % for other uses, per Table 4Ab. It is not a limit on your cable. It is a limit on everything in series in front of the load, and your cable is one term in it.

Three horizontal bars against a 3 per cent limit line. The first shows a final circuit sized on its own, using 2.30 per cent of the budget in 10 square millimetres and passing. The second shows the identical cable placed behind a sub-main that has already used 2.0 per cent: the grey upstream portion plus the same 2.30 per cent reaches 4.30 per cent and crosses the limit, although nothing about the final circuit changed. The third shows what the circuit has to become to fit inside what is left, two sizes further up the ladder. — Nothing about the final circuit changed between the first two bars. The quantity being checked is simply not a property of that row.

So the identical final circuit, unchanged in every respect, is compliant standing alone and non-compliant behind a sub-main that has already spent 2 %. Nothing about the row you are looking at tells you which. That is the honest version of "spreadsheets miss cumulative effects": not that the arithmetic is hard, but that the quantity being checked is not a property of the row. The same is true of fault level at a board, and of discrimination between a device pair, which needs both curves and the fault current at the point they meet.

There is a related trap in the withstand checks. A bigger conductor lowers Z_s, which raises the prospective fault current, which raises the energy the conductor has to survive. Sizing up attacks both sides of the adiabatic at once, so Reg 434.5.2 is not monotonic in conductor size — on a long sub-main it can be satisfied at 1.5 mm², fail at 4 mm², and settle again higher up. "The smallest size that passes" is the wrong question for that check; the right one is the smallest size above which nothing fails.

Sometimes the Answer Is Not a Cable

A bar chart of earth fault loop impedance against conductor size for a Type D 20 amp device requiring 0.4 second disconnection on a supply whose external loop impedance is already 0.80 ohms. The bars fall as the conductor grows but flatten onto the external impedance and never reach the maximum permitted loop impedance of 0.546 ohms, drawn as a limit line well below them. The supply impedance alone exceeds the limit, so no conductor can bring the loop under it. — The loop can never fall below the supply impedance. The fix is an RCD, a different device or a different supply — not a bigger cable.

A Type D 20 A device needing 0.4 s disconnection permits a maximum Z_s of about 0.55 Ω. On a supply with Z_e of 0.80 Ω, the loop is already over the limit before a single metre of cable is added, and no conductor can subtract from it. The compliant answer is an RCD, a different device characteristic, or a different supply — and a calculation that returns "no size complies" has told you something a calculation that returns a size has not. This is the case where reporting only a final cable size actively hides the problem.

What You Need to Do It

A single-line diagram or schematic (even a marked-up sketch)
The load schedule, with diversity assumptions
Cable routes and lengths, or a layout to measure from
Supply characteristics — fault level and earthing arrangement (TN-S, TN-C-S, TT)
Installation methods and grouping along the route
Preferred cable types and the protective-device manufacturer and ranges

The supply fault level is the input most often missing, and it is the one the withstand checks need. Without it the line-conductor check falls back to the earth fault current at the far end of the circuit, which is a floor rather than the worst case.

Where Spreadsheets Go Wrong

None of these checks is hard in isolation — which is exactly why a spreadsheet feels adequate. The problems are structural:

Some quantities are not per-row. Cumulative voltage drop and fault level are properties of the whole path. A spreadsheet computes one circuit at a time, so the network picture is stitched together by hand, or not at all.

Change does not propagate. When a sub-main gets longer on site, every downstream circuit's voltage drop and loop impedance move. Someone has to remember that and re-check every affected row. Most of the design errors we are asked to fix trace back to exactly this.

The governing check is invisible. A pass/fail column tells you a circuit complies, not what is holding it there — so nobody can see that forty circuits are all constrained by the same thing.

What ProDesign Automates

Trimble ProDesign (which many contractors still know by its former name, Amtech ProDesign) models the whole distribution network — supply characteristics, transformers, sub-mains, boards and final circuits — so the quantities that belong to the path are computed as path quantities. The regulation tables, correction factors and manufacturer device data with real time–current curves are built in. Change a cable length, a board location or the supply fault level, and every affected circuit is re-validated. Discrimination studies come from the same model.

It is a tool, not a warrant: the outputs are only as good as the inputs and the assumptions recorded with them, and an approving engineer is entitled to audit both. To be fair to spreadsheets, for a quick single-circuit check or an early feasibility estimate they are fine. The gap opens on real distribution networks — multiple boards, mixed installation methods, design change during construction — which is most real projects.

You can try the underlying calculation on this site: the BS 7671 cable sizing calculator runs a board of circuits through every check above, and the LV System Designer does it for a whole installation, propagating fault level and cumulative volt drop down the tree.

What a Submission-Ready Pack Contains

Whether produced in-house or outsourced, a calculation pack that gets accepted first time typically includes: the modelled single-line diagram, the design basis (supply data, demand assumptions, diversity), per-circuit results covering every check rather than a summary verdict, fault levels at each board, discrimination curves for the protective device chain, and a results summary an approving authority can review without re-running the software. Naming the governing check per circuit costs nothing and makes the whole thing reviewable.

Frequently Asked Questions

How many checks does a BS 7671 cable calculation actually make?

BS 7671 publishes no numbered list, so "the five checks" is trade shorthand rather than a citation. Counting what has to be demonstrated gives seven: overload coordination (433.1.1), current-carrying capacity (523.1), voltage drop (525.1), earth fault loop impedance (411.4.204), the protective conductor withstand (543.1.3), the line conductor withstand (434.5.2) and the minimum conductor size (524.1). Two of those share an inequality — the right-hand half of Ib ≤ In ≤ Iz is the capacity check — so there are six distinct requirements, and one of the six is a duty on the CPC rather than the line.

Which check usually decides the cable size?

On long runs it is almost always voltage drop, and often by several sizes. On short runs in hot or crowded containment it is current-carrying capacity. On sub-mains with a long earth path it is loop impedance. On short circuits behind a device that is not current-limiting, at a board with a high fault level, it is the line conductor withstand. Which one binds is a property of the job, not of the standard — and it tells you what to change: a run governed by voltage drop does not improve with a different device.

Is the installation method a derating factor?

No, and this is worth getting right. The reference method selects which column of the Appendix 4 table you read — Method C clipped direct, Method E in free air, and so on. The correction factors Ca, Cg, Ci and Cf are multipliers applied to whatever that column gives. Calling the method a factor suggests you could apply it twice, or that a change of method scales a result rather than replacing it.

Why can a cable pass on its own and fail in the installation?

Because Regulation 525.1 sets the voltage drop limit between the origin of the installation and the equipment, not across one cable. A final circuit using 2.3 % of a 3 % budget is compliant on its own and non-compliant behind a sub-main that has already spent 2 % — with nothing about the final circuit changed. The same is true of fault level and of discrimination. These are network properties, and a calculation that looks at one row at a time cannot see them.

What is Regulation 434.5.2 and why is it often missing?

It is the short-circuit withstand of the LINE conductors — the live ones — as distinct from Regulation 543.1.3, which covers the CPC. Both are the same adiabatic relationship with a different conductor and a different k. It is often left off because on most final circuits it is slack, and because for a current-limiting device clearing inside 0.1 seconds the regulation does not ask for the adiabatic at all: it asks that k²S² exceed the device’s quoted let-through energy, which is manufacturer data. That exemption names BS EN 60898-1, 60898-2 and 61009-1 — an MCCB to BS EN 60947-2 is not on the list and is judged on the adiabatic.

Is Trimble ProDesign the same as Amtech ProDesign?

Yes. Amtech ProDesign became Trimble ProDesign after Trimble acquired Amtech; it is the same calculation engine many contractors still call "Amtech". A Trimble ProDesign calculation pack satisfies a specification that asks for Amtech calculations.

What information do you need to size a cable?

A single-line diagram or schematic, the load schedule with diversity assumptions, cable routes and lengths, the supply characteristics (fault level and earthing arrangement), the installation methods and grouping along each route, and any client specification. Missing inputs are the usual cause of delay — and the supply fault level is the one most often absent, which is exactly the input the withstand checks need.

When is a spreadsheet fine, and when not?

For a quick single-circuit check or early feasibility a spreadsheet is fine; the arithmetic in any one check is not hard. It falls down on real distribution networks, because the quantities that fail are the ones that belong to the path rather than the row — cumulative voltage drop, fault level at each board, discrimination between device pairs — and because design change during construction has to propagate to every affected circuit.

What makes a calculation pack get accepted first time?

Documented inputs and assumptions, per-circuit results covering every check rather than a summary verdict, fault levels at each board, discrimination evidence for the device chain, and a results summary an approving engineer can audit without re-running the software. Saying which check governed each circuit is worth adding: it turns a table of pass marks into something a reviewer can sanity-check at a glance.

The figures and the explorer on this page are computed by the same BS 7671 engine as the site's cable sizing calculator, using an indicative subset of the Appendix 4 tables. Verify every value against your own copy of BS 7671:2018+A4:2026 before relying on it for design.

Don't have a ProDesign licence or a spare designer? We produce BS 7671-compliant calculation packs from your drawings and load schedules — see our ProDesign cable calculation service.