Discrimination & Selectivity: What BS 7671 Actually Requires
Everyone quotes a ratio. BS 7671 publishes exactly one, it lives in an informative annex, and it does not apply to circuit-breakers at all. Here is what the standard actually says, where the number really comes from, and what to do for the pairings it does not answer.
A fault on one lighting circuit should trip that circuit’s breaker — not the incomer feeding the whole building. Keeping a fault contained to the device nearest it is selectivity. It is a design outcome you have to prove, and it is routinely assumed instead.
The assumption usually rests on a rating ratio: step the devices 1.6:1, or 2:1, and you are covered. That belief has a real origin, and tracing it back is the most useful thing this article can do — because the rule is genuine, it is in BS 7671, and it applies to almost none of the situations it gets used in.
1. What Section 536 actually obliges
Section 536 is titled co-ordination of protective devices, and the first surprise is what is missing from it: Section 536 does not require selectivity. It tells you how to verify selectivity “where selectivity is required” — and the requirement itself comes from somewhere else. A life-safety or standby system, a fire alarm or smoke control standard, or the client’s specification. Full discrimination between every device on every board is not a blanket BS 7671 duty.
Where it is required, Regulations 536.4.1.2, 536.4.1.3 and 536.4.1.4 each give the same four routes, almost word for word: a desk study taking account of the relevant product standard and the manufacturer’s literature; appropriate software tools where the manufacturer provides information for that specific use; tests to the applicable product standard; or a manufacturer’s declaration. None of the four is arithmetic on a rating ratio.
The desk-study route carries two conditions worth knowing, because it is the route most people actually take. Reg 536.4.1.2.1 requires you to account for the reference ambient temperature the tripping curves are drawn for, and for the load conditions before the overcurrent — a warm board and a pre-loaded bimetal both move the curve. Reg 536.4.1.3.1 adds that energy-limitation curves must be read at the voltage they are given for.
| Regulation | Subject | What it says |
|---|---|---|
| 536.4.1 | General | Signposts the rest. There is no Regulation 536.4.1.1 — the numbering runs 536.4.1 straight to 536.4.1.2. |
| 536.4.1.2 | Selectivity under overload, between OCPDs | Where selectivity is required, verify it by one of four routes. A desk study must account for the ambient temperature the curves are drawn for and the load before the overcurrent. |
| 536.4.1.3 | Selectivity under short circuit, between OCPDs | The same four routes. Energy-limitation curves must be read at the voltage they are given for. Selectivity may be total or partial; the selectivity limit current comes from the manufacturer. |
| 536.4.1.4 | Selectivity between RCDs | The one numerically specific clause. Upstream must be a selective (type S) or time-delayed device and at least 3× the downstream IΔn. |
| 536.4.1.5 | RCDs and an upstream OCPD | An earth fault can exceed the upstream device's instantaneous trip. Use an RCBO or CBR — with RCCBs, selectivity cannot be guaranteed. |
| 536.4.1.6 | RCDs and a downstream OCPD | The mirror image: the fault can be below the downstream device's instantaneous trip, so the upstream RCD goes first. Use RCBOs. |
| 536.4.2.1 | Combined short-circuit protection | The permission that inverts the intuition — and the rule that where the manufacturer gives no information, it shall not be used. |
Note the gap. Guidance that cites “Regulation 536.4.1.1” is citing something that does not exist — the numbering runs 536.4.1 straight to 536.4.1.2.
2. Where the ratio rule really comes from
The number is real and it is in the standard. It is in Annex A53, which is informative, and it appears in exactly two clauses:
- A53.4 (overload): fuses to BS EN 60269-2 of the same utilization category — type gG for example — with a rated current of 16 A and above will provide total selectivity if the ratio of the rated currents is 1.6:1 or greater.
- A53.8 (short circuit): the same 1.6:1, for fuses to BS EN 60269-1, with selectivity verified by comparing I²t values.
Three qualifiers travel with that number and are almost always dropped in the retelling. It is fuse to fuse only. Both links must be of the same utilization category. And it does not apply below 16 A. The annex is also informative — a recognised way of demonstrating compliance, not a requirement in itself.
Why does it work for fuses? Because a fuse-link has one mechanism and one curve shape. Its pre-arcing and total I²t scale predictably with rated current, so a step in rating really does separate the characteristics. A53.4 states the underlying test in words: the total operating time of the load-side fuse must be lower than the pre-arcing time of the supply-side fuse. The ratio is a shortcut to that comparison, not a substitute for it — A53.8 says the short-circuit case is verified by comparing I²t.
And why does it not transfer? A circuit-breaker has two mechanisms. In the thermal region a rating step behaves much as it does for a fuse. Above the magnetic threshold both devices operate in a few milliseconds regardless of how far apart their ratings are, because neither is limited by anything the ratio describes. That is why A53.7 gives no number at all and opens by saying to refer to the manufacturers’ documentation.
| Clause | Condition | Pairing | Method | Ratio |
|---|---|---|---|---|
| A53.3 | Overload | Breaker over breaker | Downstream maximum operating time below the upstream non-tripping time. | — |
| A53.4 | Overload | Fuse over fuse | Downstream total operating time below the upstream pre-arcing time. | 1.6:1 |
| A53.5 | Overload | Breaker over fuse | Maximum operating time curve for the downstream device, minimum for the upstream. | — |
| A53.6 | Overload | Fuse over breaker | Breaker maximum operating time below the fuse minimum pre-arcing time. | — |
| A53.7 | Short circuit | Breaker over breaker | Selective up to the level at which the downstream peak let-through is below the upstream instantaneous level. Refer to the manufacturer. | — |
| A53.8 | Short circuit | Fuse over fuse | Downstream maximum operating I²t below the upstream minimum pre-arcing I²t. | 1.6:1 |
Two clauses out of six carry a number, and both of them are fuse over fuse.
3. The trip is a band, not a point
A BS EN 60898 circuit-breaker’s instantaneous trip is specified as a range: Type B operates somewhere between 3 and 5 times rated current, Type C between 5 and 10, and Type D between 10 and 20. Which end you use depends entirely on the question you are asking, and using the wrong one is a quiet, common error.
For a disconnection-time check you need the current at which the device will
definitely have operated — the late edge. BS 7671 states it directly: Table 41.3
gives the maximum earth fault loop impedance as
Zs = 230 × 0.95 / (5·In) for Type B,
/(10·In) for Type C and /(20·In)
for Type D. Those denominators are Ia.
For selectivity you need the opposite: the current at which the upstream device might already have started to operate. That is the early edge — 3, 5 and 10 times In. Take the late edge here and you will call a pair selective well past the point it stops being so.
Annex A53.5 states the same principle in terms of whole curves, and it is the single most useful sentence in the annex: when using published time/current characteristics, take the maximum operating time curve for the downstream device and the minimum operating time curve for the upstream one. The pessimistic edge of each device is a different edge. A bare rating ratio implicitly compares two nominal centre-lines, which is optimistic on both sides at once.
This is also why two MCBs in series discriminate so poorly. Below the upstream device’s magnetic pickup the upstream breaker is still in its slow thermal region while the downstream one clears magnetically in milliseconds, so the downstream device clears alone. Above it, both are in their instantaneous region and the rating step stops mattering. Selectivity between MCBs is therefore normally partial, and the honest design question is whether the prospective fault at that board sits below the limit.
4. RCDs — the one place BS 7671 is numeric
Regulation 536.4.1.4 is the exception to everything above, and it is specific. Selectivity between RCDs in series is given under both of the following conditions:
- the upstream RCD is of selective type — type S, or a time-delayed type with an appropriate delay setting; and
- the ratio of the upstream device’s rated residual operating current to the downstream device’s is at least 3:1.
Both, not either. A 100 mA device over a 30 mA device is 3.3:1 and still fails the regulation if it is not an S type — and a plain 100 mA RCD in series with a 30 mA one is exactly the arrangement people reach for. NOTE 4 adds a condition that is easy to miss: selectivity holds for both a line-to-earth and a neutral-to-earth fault only where the downstream RCD switches all live conductors, including the neutral.
Regulations 536.4.1.5 and 536.4.1.6 handle the mixed case, and both land on the same answer. Feeding downstream OCPDs from an RCCB cannot guarantee selectivity, because an earth fault may be above the upstream device’s instantaneous trip or below the downstream one’s. Where selectivity is required between an RCD and an OCPD, BS 7671 says to use RCBOs.
5. The mirror image: a device below the fault current
If Section 536 gives less help than expected in one direction, it gives more permission than expected in the other. A device whose breaking capacity is lower than the prospective fault current at its own terminals can be entirely legitimate.
The route through the standard is not the one most guidance gives. Regulation 533.3 permits the arrangement where a suitable device is installed on the supply side. Regulation 434.5.1 carries the criterion: the two devices must be co-ordinated so the energy let through by the upstream device does not exceed what the downstream device and its conductors can withstand without damage. And Regulation 536.4.2.1 is the operative requirement — refer to the downstream device manufacturer’s instructions, which are derived from type tests, and where no such information is available, combined short-circuit protection shall not be used and each device shall have the required capability on its own.
One thing to know if you follow the references yourself: 533.3 points at “the last paragraph of Regulation 536.1”, and that paragraph is about continuity of supply. It does not contain the criterion. Follow the cross-reference literally and you will conclude the permission is unqualified; it is not. The conditions are in 434.5.1 and 536.4.2.1.
The everyday example is the domestic consumer unit with a conditional short-circuit rating, conditional on the distributor’s fuse ahead of it. It is a genuine, tested arrangement. What it is not is selectivity: it works precisely because the upstream device helps clear the fault, which means both devices may operate. A cascade table and a selectivity table answer opposite questions.
6. The words, and which ones BS 7671 has
A surprising amount of confusion here is vocabulary. Three terms in common use do not appear in BS 7671 at all, and two that do appear mean different things and are routinely swapped.
| Term | In BS 7671? | What it means |
|---|---|---|
| Selectivity | yes | Only the device nearest the fault operates. BS 7671 uses this word. |
| Discrimination | cross-reference | Part 2 lists it only as “Discrimination (see Selectivity)”. Same meaning; the standard has moved to the international term. |
| Total selectivity | yes | Holds up to the breaking capacity of the downstream device. |
| Partial selectivity | yes | Holds only up to the selectivity limit current. Above it, both devices may operate. |
| Combined short-circuit protection | yes, 536.4.2.1 | Two OCPDs, the upstream one limiting let-through so the downstream may sit below the fault current at its terminals. |
| Back-up protection | yes, 536.4.2.2 | Not the same thing. In BS 7671 this is protecting a device that is not an OCPD — a contactor, an overload relay. |
| Cascading | no | Appears nowhere in BS 7671. It is the trade word for combined short-circuit protection. |
| Zone-selective interlocking | no | A real manufacturer feature, but BS 7671 does not mention it. |
| Icu / Ics / Icn | no | BS EN 60947-2 and 60898-1 symbols. BS 7671 writes the capacity out in words, and Part 2 has no definition of “breaking capacity” at all. |
BS 7671 states outright that back-up protection is not the same as combined short-circuit protection. Reading a cascade table as a discrimination table is the classic consequence.
7. So what do you actually do?
- Fuse over fuse. Apply Annex A53.4 / A53.8: same utilization category, both 16 A and above, ratio 1.6:1 or greater. Confirm against the published pre-arcing and total I²t for the actual fuse-links.
- Breaker over breaker. There is no ratio. Establish the upstream device’s magnetic pickup, compare it with the prospective fault at that board, and get the manufacturer’s tested selectivity table for the pair — that is the manufacturer’s-declaration route in 536.4.1.3.
- MCCB over MCB. The variable that decides it is whether the MCCB has an adjustable short-time element. Raise the short-time pickup above the downstream let-through, and a short-time delay extends selectivity above it — up to any instantaneous override, which cancels the delay.
- Mixed fuse and breaker. No ratio transfers. A53.5 and A53.6 give the curve comparison, each naming which edge to take.
- RCDs. Reg 536.4.1.4 — S type and 3:1, and check the downstream device switches the neutral.
- Record which of the four routes you used. “Verified by desk study against the manufacturer’s published curves” is a defensible statement. “The ratio is 2:1” is not.
Frequently asked questions
Does BS 7671 give a selectivity ratio for circuit-breakers?
No. Section 536 contains no numeric selectivity ratio of any kind, and the informative Annex A53 gives a figure only for fuses. A53.7, which covers circuit-breaker pairs, states the principle — selectivity holds up to the fault level at which the downstream device's peak let-through stays below the upstream device's instantaneous tripping level — and then says to refer to the manufacturers' documentation. The widely quoted "2:1 rule" for MCBs is not in the standard.
Where does the 1.6:1 rule actually come from?
BS 7671 Annex A53, clauses A53.4 (overload) and A53.8 (short circuit). Both state that fuses to BS EN 60269 of the same utilization category — type gG for example — with a rated current of 16 A and above give total selectivity where the ratio of rated currents is 1.6:1 or greater. Three qualifiers matter and are usually dropped: it is fuse-to-fuse only, it needs the same utilization category, and it does not apply below 16 A. Annex A53 is also informative, not a requirement.
What is the difference between total and partial selectivity?
Total selectivity holds up to the breaking capacity of the downstream device. Partial selectivity holds only up to a lower current — the selectivity limit current, which the manufacturer states — and above that point the upstream device may operate too. Both terms are defined in Part 2 of BS 7671, and Regulation 536.4.1.3 says explicitly that in a particular installation selectivity may be total or partial.
Why can two MCBs in series both trip?
Above the magnetic threshold an MCB clears in a few milliseconds through its instantaneous element. Two MCBs in series see essentially the same fault current, so once that current exceeds the upstream device's magnetic pickup both devices start to operate and the rating step between them becomes irrelevant. Time grading cannot help either, because neither device has a deliberate delay. Below that threshold the upstream device is still in its slow thermal region and the downstream one clears alone, which is why selectivity between MCBs is normally partial.
Does BS 7671 require selectivity?
Section 536 does not impose a blanket requirement. It tells you how to verify selectivity "where selectivity is required" — the requirement itself comes from elsewhere, typically a life-safety or standby system, another standard, or the client specification. Where it is required, Regulations 536.4.1.2, 536.4.1.3 and 536.4.1.4 each give the same four verification routes: a desk study, appropriate software tools, tests to the product standard, or a manufacturer's declaration.
What does Regulation 536.4.1.4 require for RCDs in series?
Both conditions together, not either one. The upstream RCD must be a selective type (type S) or a time-delayed type with an appropriate delay setting, and the ratio of the upstream device's rated residual operating current to the downstream device's must be at least 3:1. A 100 mA non-delayed device over a 30 mA device clears the ratio and still fails, because it is not an S type. NOTE 4 adds a condition people miss: selectivity holds for both line-to-earth and neutral-to-earth faults only where the downstream RCD switches all live conductors, including the neutral.
Can a device have a breaking capacity lower than the fault current at its terminals?
Yes, and it is a recognised technique. Regulation 533.3 permits it where a suitable device is installed on the supply side, Regulation 434.5.1 gives the criterion — the energy let through by the upstream device must not exceed what the downstream device and its conductors can withstand without damage — and Regulation 536.4.2.1 makes it conditional on the downstream device manufacturer's instructions, which are derived from type tests. Where no such information is available, BS 7671 says combined short-circuit protection shall not be used and each device shall have the required capability on its own.
Is cascading the same as discrimination?
No, and they pull in opposite directions. Combined short-circuit protection — what the trade calls cascading — works precisely because the upstream device helps clear the fault, which means both devices may operate. Selectivity means only the nearest one operates. A cascade table and a selectivity table answer different questions and reading one as the other is a common error. Note also that "cascading" is not a BS 7671 term, and that the standard reserves "back-up protection" for something different again: protecting a device that is not an overcurrent protective device, such as a contactor.
Which end of the trip band should I use?
It depends on the question. For a disconnection-time or maximum-Zs check you need the current at which the device will definitely have operated, which is the late edge of the band — Table 41.3 states it directly as Ia = 5·In for Type B, 10·In for Type C and 20·In for Type D. For selectivity you need the current at which the upstream device might already have started to operate, which is the early edge: 3, 5 and 10 times In respectively. Annex A53.5 puts the same idea in terms of curves: take the maximum operating time curve for the downstream device and the minimum operating time curve for the upstream one.
The curves and thresholds shown here are generic band shapes derived from the BS EN device classes, not manufacturer tested data. Annex A53 is informative. Every real device pair must be confirmed against the manufacturer’s published time/current characteristics and tested selectivity tables before it is relied upon.
Where this fits
Selectivity is decided by the prospective fault current at each board, so it sits directly on top of the fault-level calculation — see prospective short-circuit current for how that figure is derived, and RCD types for the residual-current side. The selectivity checker runs the same engine as the widget above, and the fault level calculator gives you the current to check it against.
Getting this right across a whole installation is exactly the kind of thing outsourced design is for. Get in touch if you would like it done properly, or see ProDesign cable calculations for how it fits into a full design package.