RCD Types Explained: AC, A, F and B — Which Do You Need?

An RCD works by summing the current going out on the line against the current coming back on the neutral. Any difference is current that left the circuit somewhere it should not have — the residual current. That much is well known.

What is much less well known is that the device does not respond to residual current in general. It responds to particular shapes of it.

The short version: a Type AC trips on alternating sinusoidal residual current and nothing else. Feed a few milliamps of smooth DC through its core and it is desensitised, and can stop responding altogether — BS 7671 calls this RCD blinding. Almost every modern load produces exactly the kind of non-sinusoidal leakage that does it. That physical fact is why Reg 531.3.3 restricts Type AC to a set of loads so narrow that in practice nothing on a modern job qualifies.

This article is written against BS 7671:2018+A4:2026. Where the amendment history matters — and on this subject it is widely misreported — it says which amendment actually did what.

1. The shape of the fault current is the whole problem

A residual current is not automatically a sine wave. What it looks like depends on what the circuit feeds. A plain insulation failure on a resistive load gives a sinusoid. A rectifier front end — which is the input stage of essentially every switch-mode power supply, LED driver and charger — gives a pulsating DC. An inverter drive gives a composite of mixed frequencies. A drive with a smoothed DC link gives something that is, to the sensing core, very nearly steady DC.

Four residual-current waveforms with the RCD types that respond to each. A sinusoidal AC fault is seen by Types AC, A, F and B. Pulsating DC from a rectifier front end is seen by A, F and B but not AC. Composite mixed-frequency current from an inverter drive is seen by F and B. Smooth DC is seen only by Type B. — A Type AC responds to the top row and nothing else. Almost no modern circuit is the top row.

A current transformer works on changing flux. A steady DC component does not change, so it produces no output — but it does bias the core, pushing its operating point toward saturation. Once there, the same alternating residual current that would have tripped the device produces far less change of flux than it should.

Left, the sensing core with line and neutral passing through it, summing the two so any difference is current leaking to earth. Right, a magnetisation curve showing that a smooth DC component shifts the operating point from the responsive linear region into saturation, where the same AC residual current produces far less change of flux. — The device is still fitted, still looks fine, and has stopped responding. BS 7671 owns the term: Figure A53.2 NOTE 5 calls it RCD blinding.

The device is still fitted. It still looks fine. Its test button may still work, because the test button injects a current through a dedicated winding rather than reproducing the fault. And it has stopped protecting the circuit. Annex A53, Figure A53.2, NOTE 5 gives it a name: "Impaired operation is commonly termed as 'RCD blinding'."

2. The four types (Reg 531.3.3)

Regulation 531.3.3 opens: "Different types of RCD exist, depending on their behaviour in the presence of DC components and frequencies." The four types are its lettered indents — there is no Regulation 531.3.3.1, and there is no definition of any RCD type in Part 2. These four indents are the only place BS 7671 defines them.

Reg 531.3.3 — the four types, as printed
Type	Reg	Trips on	Smooth DC it tolerates	Product standard
AC	531.3.3(a)	Alternating sinusoidal residual current.	—	BS EN 61008 / 61009
A	531.3.3(b)	As AC, plus residual pulsating DC.	6 mA	BS EN 61008 / 61009
F	531.3.3(c)	As A, plus composite residual currents on a line-to-neutral circuit.	10 mA	BS EN 62423
B	531.3.3(d)	As F, plus residual smooth DC of either polarity, and sinusoidal AC to 1 kHz.	0.4 IΔn or 10 mA, whichever is higher	BS EN 62423

Read that fourth column carefully, because it is the most misquoted thing in this subject. The 6 mA against Type A is not smooth DC a Type A can detect. It is the smooth DC a Type A tolerates while still tripping on a pulsating component superimposed on it. NOTE 1 to the regulation says so in as many words: "tripping is achieved for residual pulsating direct currents superimposed on a smooth direct current up to 6 mA". Only a Type B responds to smooth residual DC — that is indent (d)(v).

Note also who defines these types. It is not BS 7671. Regulation 531.3.4.1 points at the product standards: BS EN 61008 and BS EN 61009 for Types AC and A, and BS EN 62423 for Types F and B. BS 7671 selects among them; it did not create them.

3. Why Type AC is effectively finished

Sitting between NOTE 4 and NOTE 5 of Regulation 531.3.3 is a single unnumbered sentence that does most of the work:

RCD Type AC shall only be used to serve fixed equipment,
where it is known that the load current contains no DC components.

Three things about that. First, it is a restriction, not a ban — a Type AC is still permitted for the loads it names. Second, NOTE 5 tells you what those are: "electric heating appliances and/or simple filament lighting, neither containing electronic components." Ask yourself how much of a modern installation that describes. Third — and this is where a great deal of published guidance is simply wrong — this landed in Amendment 2:2022.

On the amendment history, since it is so often garbled: the introduction to Amendment 2:2022 states the change directly. Amendment 3:2024 changed nothing about RCDs at all — its entire technical scope was "two new definitions and one new Regulation relating to the use of bidirectional and unidirectional devices". Amendment 4:2026 did not change the type definitions either; its work in this area was on the product-standard lists in 531.3.4 and 531.3.6, the guidance NOTEs, and the Annex A53 figures.

There is exactly one place in the whole standard where a type is outright prohibited: Regulation 710.531.3.3.101 — "In medical locations of group 1 and group 2, RCDs of Type AC shall not be used."

4. The standard selects on topology, not on appliance

The most useful thing in this subject is buried in an annex. Annex A53, Figure A53.1 — "Possible fault currents in systems with semiconductors" — maps nine circuit topologies to the RCD types that give protection for each. Regulation 531.3.3 NOTE 7 points at it, and almost nobody reads it.

The standard selects the type from the CIRCUIT, not from the appliance name. Four of the nine topologies leave Type B as the only answer.

Notice what it is keyed on. Not "heat pump" or "EV charger" or "washing machine", but the rectifier topology at the front of the equipment. That is the honest basis for a selection, and it is also the awkward one, because the topology is not usually printed on the box. In practice you read it off the manufacturer's stated RCD requirement — which is why every regulation in this area ends up pointing back at the manufacturer's instructions.

The headline is the bottom four rows. A single-phase circuit with smoothing, a two-pulse bridge between phases, a three-phase star and a six-pulse bridge all produce an essentially smooth DC earth fault current, and for all four Type B is the only answer.

5. Where BS 7671 requires 30 mA

The type and the rating are two separate decisions, and the regulations that demand a 30 mA device are almost entirely silent about type.

Five requirements for 30 mA RCD protection: Regulation 411.3.3 for socket-outlets rated 32 A or less in any location, with a documented risk-assessment exception that applies to other locations only; 411.3.4 for luminaire final circuits in domestic premises; 701.411.3.3 for bathrooms; 722.531.3.101 for every EV charging point individually; and 708.415.1 for caravan park and marina socket-outlets individually. — None of these regulations names an RCD type. The 30 mA rating and the type are two separate decisions.

Regulation 411.3.3 requires 30 mA additional protection for socket-outlets rated 32 A or less — in locations used by ordinary persons or children (indent a), in other locations (indent b), and for mobile equipment up to 32 A used outdoors (indent c). There is an exception, and its shape matters: it applies to (b) but not (a) or (c), it needs a documented risk assessment undertaken with the involvement of a skilled person, and that assessment must be issued with the certificate. NOTE 3 to the regulation adds, drily, that RCD protection of all socket-outlets is recommended anyway.

Regulation 411.3.4 is two lines long and is missed constantly: "Within domestic (household) premises, additional protection by an RCD with a rated residual operating current not exceeding 30 mA shall be provided for AC final circuits supplying luminaires." Note the scope difference — 411.3.3 applies everywhere, 411.3.4 only in dwellings. This is not a recent addition: it was new in BS 7671:2018 itself.

And Regulation 415.1.2 is worth keeping in view: the use of RCDs "is not recognized as a sole means of protection and does not obviate the need to apply one of the protective measures specified in Sections 411 to 414." Additional protection is additional to a design that already works.

6. EV charging, in full

This is the question that arrives most often and the one most often answered loosely. Regulation 722.531.3.101 says that, unless the circuit uses electrical separation, each charging point incorporating a socket-outlet or vehicle connector to the BS EN 62196 series shall be protected individually by an RCD of Type A, Type F, Type B or an RDC-PD to BS IEC 62955, with a rated residual operating current not exceeding 30 mA.

Then, separately: except where provided by the EV charging equipment, protection against DC fault currents shall be provided by

(a) an RCD Type B; or
(b) an RCD Type A or Type F in conjunction with a residual direct current detecting device (RDC-DD) conforming to BS IEC 62955; or
(c) an RDC-PD conforming to BS IEC 62955.

Four details people miss. Type F is an option, not just A or B. The protection must be individual per charging point — one RCD each, and Regulation 722.533.101 separately requires each charging point to be supplied by its own final circuit. NOTE 3 confirms the arrangement most chargers actually use: the RDC-DD can sit inside the charging equipment with the Type A or F upstream. And the familiar 6 mA figure does not appear in Section 722 at all — it is in the Part 2 definitions of the RDC-DD and RDC-PD, both of which are defined in terms of detecting 6 mA DC residual currents.

7. RCDs in series — the trap in an upgrade

Annex A53 Figure A53.2 deals with what happens when RCDs of different types sit in series, and it is the part of this subject with the least coverage anywhere.

An upstream RCD feeding four downstream devices of Types AC, A, F and B, showing which combinations are a blinding risk. A lower type upstream of a higher one can have its own operation impaired by the load the downstream device was selected for, unless the manufacturer specifically declares the arrangement suitable. — Figure A53.2. Getting the final circuit right and leaving an older device upstream can blind the upstream one.

The rule runs the way round that surprises people. It is not the downstream device that is at risk — it is the upstream one. A Type AC should not sit upstream of a Type A, F or B; a Type A should not sit upstream of a Type F or B; a Type F should not sit upstream of a Type B. In each case the load characteristics the downstream device was chosen for are exactly the characteristics that can impair the upstream one.

The practical trap: you fit a Type B for a new EV charge point or a heat pump, and leave the existing Type AC or Type A main switch RCD at the origin. The new circuit is correctly protected. The old device upstream, which is still protecting everything else, is now exposed to precisely the DC content it cannot handle. NOTE 4 permits the arrangement only where the RCD manufacturer specifically declares it suitable — so this is a question to put to the manufacturer, not one to reason your way past.

8. Type is not rating — and the rating is not the trip current

The rated residual operating current IΔn is a separate axis from the type. A 30 mA Type A and a 300 mA Type A are both "Type A" and do entirely different jobs.

It is also not the current at which the device trips. Regulation 531.3.2 NOTE 2: "RCDs may operate at any value of residual current in excess of 50 % of the rated residual current." A 30 mA device is permitted to operate anywhere above 15 mA — which is exactly why the next rule exists.

Regulation 531.3.2(c): "in order to avoid unwanted tripping by protective conductor currents and/or earth leakage currents, the accumulation of such currents downstream of the RCD shall be not more than 30 % of the rated residual operating current." On a 30 mA device that is 9 mA of standing leakage, total, for everything downstream. It is not a large budget once a board is full of electronics, and it is the real reason Regulation 531.3.2(b) points at RCBOs on individual final circuits in dwellings rather than one RCD across a whole board.

Separately, where a single item of equipment has a protective conductor current exceeding 10 mA, Regulations 543.7.1.202 and 543.7.1.203 bring in high-integrity protective conductor arrangements — a different requirement, often confused with this one.

For ratings above 30 mA: 300 mA or less is the figure for protection against fire (Reg 532.2), and where RCDs are in series Regulation 536.4.1.4 wants the upstream device to be selective or S type and at least three times the downstream rating. Both conditions, not either.

9. TT systems: a table, not just a formula

On a TT system the RCD is doing fault protection, not just additional protection, and Regulation 411.5.3 gives the familiar condition R_A × IΔn ≤ 50 V. What is less familiar is that Regulation 531.3.5.3.2 turns it into a table.

Table 53.1 — maximum earth resistance R_A against the maximum IΔn
Maximum R_A (Ω)	Maximum IΔn of the RCD
2.5	20 A
5	10 A
10	5 A
17	3 A
50	1 A
100	500 mA
167	300 mA
500	100 mA
1667	30 mA

The companion is Table 41.5, which gives the maximum earth fault loop impedance for an RCD at U₀ 230 V — and has only four rows: 30 mA → 1667 Ω, 100 mA → 500 Ω, 300 mA → 167 Ω, 500 mA → 100 Ω. Read the note attached to the first two: an electrode resistance above 200 Ω may not be stable (Reg 542.2.4). The 1667 Ω figure is arithmetic, not a design target. Our earthing systems explainer covers the TT loop in detail.

One more, easily forgotten: Regulation 531.3.5.2 — "An RCD shall not be used in a TN-C system." There is no separate protective conductor for an imbalance to be measured against.

10. Testing — the five-times test is gone

If you learned RCD testing before 2022, the procedure has changed. Amendment 2:2022 deleted Appendix 3 Table 3A, which held the time/current performance criteria, and replaced the sequence with a single alternating-current test at the rated residual operating current. Regulations 643.7.1 and 643.8 put it plainly: "Regardless of RCD Type, effectiveness is deemed to have been verified where an RCD disconnects within the time stated below with an alternating current test at rated residual operating current" —

general non-delay type: 300 ms maximum;
delay 'S' type: between 130 ms minimum and 500 ms maximum.

Appendix 3 now covers overcurrent devices only. Note the phrase "regardless of RCD Type" — the test does not attempt to prove the device's type. That is a matter of selection and record-keeping, which is why the Appendix 6 model forms now carry an RCD Type AC / A / F / B field on the schedule of circuit details.

Regulation 514.12.2 still requires the six-monthly test-button notice at or near the RCD, and Regulation 643.10 requires the device's own test facility to be verified.

Glossary

Residual current: The algebraic sum of the currents in the live conductors of a circuit at a point in the installation. If it is not zero, current is leaving the circuit.
IΔn: Rated residual operating current. The value at which the device must have operated — not where it trips, which may be anywhere above 50 % of it.
RCCB / RCBO: A residual current operated circuit-breaker without integral overcurrent protection (RCCB), and with it (RCBO).
RDC-DD: Residual direct current detecting device: detects and evaluates 6 mA DC residual current and switches the monitored circuit. BS IEC 62955.
RDC-PD: Residual direct current protective device: integrated AC, pulsating DC and 6 mA DC detection with mechanical switching.
Blinding: Impaired operation of an RCD caused by residual current it cannot respond to. The term is BS 7671's own — Annex A53, Figure A53.2, NOTE 5.
Additional protection: Protection in the event of failure of basic or fault protection, or carelessness by users. By RCD it means 30 mA or less (Reg 415.1.1). BS 7671 has no Part 2 definition of the term itself.

Frequently Asked Questions

What is the minimum RCD type for a new installation?

In practice Type A. BS 7671 does not say so in those words — what Regulation 531.3.3 says is that a Type AC "shall only be used to serve fixed equipment, where it is known that the load current contains no DC components", and its NOTE 5 gives the examples: electric heating appliances and simple filament lighting, neither containing electronic components. Almost nothing on a modern installation qualifies. So Type A is the floor for general work, and Figure A53.1 pushes several common circuit types straight past it.

Is a Type AC RCD banned?

Not in general. Regulation 531.3.3 restricts it rather than prohibiting it, which is a different thing — it may still serve fixed equipment whose load current contains no DC component. There is exactly one outright prohibition in the whole standard: Regulation 710.531.3.3.101 says that in medical locations of group 1 and group 2, RCDs of Type AC shall not be used.

Which amendment restricted Type AC?

Amendment 2:2022. Its own introduction says it plainly: "Regulation 531.3.3 now states that RCD Type AC shall only be used to serve fixed equipment, where it is known that the load current contains no DC components." A lot of published guidance attributes this to a later amendment, which is wrong — Amendment 3:2024 was a very small amendment whose entire technical scope was two new definitions and one new regulation about bidirectional and unidirectional devices. It changed nothing about RCDs.

What is the difference between Type A, Type F and Type B?

Each is a superset of the one before. A Type A adds residual pulsating DC to what a Type AC does. A Type F adds composite mixed-frequency residual currents for a circuit supplied line-to-neutral. A Type B adds, among other things, residual smooth DC of either polarity and sinusoidal AC up to 1 kHz. The published numbers — 6 mA for Type A, 10 mA for Type F, and 0.4 times the rated residual current or 10 mA whichever is higher for Type B — are the smooth DC each type TOLERATES while still tripping on a superimposed pulsating component. They are not smooth DC the device can detect. Only a Type B detects smooth residual DC.

What RCD does an EV charge point need?

Regulation 722.531.3.101 is specific. Each charging point with a socket-outlet or vehicle connector to the BS EN 62196 series shall be protected INDIVIDUALLY by an RCD of Type A, Type F, Type B or an RDC-PD to BS IEC 62955, with a rated residual operating current not exceeding 30 mA. Then, except where the charging equipment provides it, protection against DC fault currents shall be by a Type B; or by a Type A or Type F together with a residual direct current detecting device (RDC-DD) to BS IEC 62955; or by an RDC-PD. Note that Type F is an option, that the protection must be individual per charge point, and that the familiar 6 mA figure is not in Section 722 at all — it is in the Part 2 definitions of the RDC-DD and RDC-PD.

Do solar PV and heat pumps need a Type B?

PV often does. Regulation 712.531.3.5.1 requires a Type B on the PV AC supply circuit unless the inverter provides at least simple separation between its AC and DC sides, or the installation does so through separate transformer windings, or the inverter manufacturer states a Type B is not required. Heat pumps are not named by any regulation — the answer depends on the drive topology inside, which is exactly what Annex A53 Figure A53.1 is for, and on what the manufacturer states. A single-phase inverter with a smoothed DC link is row 6 of that figure, and row 6 is Type B only.

Can one RCD blind another?

Yes, and BS 7671 owns the term. Annex A53 Figure A53.2 warns that a Type AC should not sit upstream of a Type A, F or B, a Type A should not sit upstream of a Type F or B, and a Type F should not sit upstream of a Type B, because the load characteristics the downstream device was selected for could impair operation of the upstream one. NOTE 5 to that figure says: "Impaired operation is commonly termed as RCD blinding." NOTE 4 permits the arrangement where the RCD manufacturer specifically declares it suitable. The practical trap is upgrading a final circuit to a Type B and leaving an older device at the origin.

What is the difference between a 30 mA and a 100 mA RCD?

The rated residual operating current is separate from the type, and it is not the current at which the device trips — Regulation 531.3.2 NOTE 2 permits an RCD to operate at any residual current above 50 per cent of its rated value. Only 30 mA or less counts as additional protection against electric shock under Regulations 415.1.1 and 531.3.6. Larger ratings do other jobs: 300 mA or less for protection against fire under Regulation 532.2, and higher ratings upstream for selectivity, where Regulation 536.4.1.4 wants the upstream device to be a selective or S type and at least three times the downstream rating.

Is the 5 times rated-current RCD test still required?

No. Amendment 2:2022 deleted Appendix 3 Table 3A, which held the time/current performance criteria for RCDs, and replaced the whole thing with a single alternating-current test at the rated residual operating current. Regulations 643.7.1 and 643.8 now say that regardless of RCD type, effectiveness is verified where the device disconnects within 300 ms for a general non-delay type, or between 130 ms and 500 ms for a delay S type. Appendix 3 now covers overcurrent devices only.

Where this fits

RCD selection sits inside the protection design for a board, next to surge protection — which has its own RCD interaction, since an SPD downstream of an RCD needs one with at least 3 kA 8/20 surge immunity — and discrimination. The earthing arrangement decides how much work the RCD is doing in the first place, which is covered in TN-S, TN-C-S and TT explained. You can check a circuit's disconnection numbers in our free BS 7671 cable sizing calculator, and we fold RCD and SPD selection across a whole distribution system into a full design package.

Indicative, for learning — not a design. Every regulation reference here is transcribed from BS 7671:2018+A4:2026 as an aid to understanding it, not as a substitute. The circuit topologies come from Annex A53 Figure A53.1; the plain-English descriptions of what uses each are ours, not the standard's. Confirm the equipment's actual front end and the manufacturer's stated RCD requirement, and check every figure against your own copy of BS 7671 before you specify anything.