Ring Final Circuit (RFC) versus Radial. Yes, again.

[pc1966]

This is a topic that comes up a lot and seems to attract near-religious levels of fever in how folks argue over the choice of circuit for UK sockets. However, often the arguments are ill-informed or they veer off in to pointless corners, or sometimes get personal. This is my attempt to put a lot of the information and arguments in to one place, in a couple of hopefully readable posts, so at least future debates are better informed. So which is best?

TL;DR Neither, they both work well in certain cases:

• If you have a flat or floor of a house then probably the RFC is the best choice.
• If you have a group of sockets in one place probably the radial is the best choice

In both cases it is “probably” as you can easily think of cases where a few radials are better than RFC(s), or an RFC better for a few sockets. The key to choosing is knowledge of the how and why. More posts to follow.

 

[pc1966]

UK versus The World

One false argument used against the RFC is why is it not used elsewhere in the world?

TL;DR The UK has fused plugs, the rest [*] don't

The UK uses BS 1363 standard plugs and sockets for virtually all domestic and most commercial single-phase outlets, these have a few features that are very important to both the design of the socket circuits and to the testing:

• Fused: Each plug has a HRC fuse, typically from 3A to 13A, in the line conductor that is appropriate for the attached flex and appliance.
• Polarised: The design is such that polarity is enforced, important due to the line-only fusing (and often switching).
• Earth: All sockets support an earth pin that makes first and breaks last, for ADS (Automatic Disconnection of Supply), as a means to polarise orientation, and to operate the safety shutters on the live pins.

When it comes to the design of UK “final” socket circuits you could argue that due to the fuses they are not really final, they are actually more of a distribution circuit to a set of variable OCPD, each related to what is plugged in. This aspect is fundamental to the UK approach and very different to the rest of the world as it separates the OCPD in the CU that protects the fixed wiring, from the OCPD needed to protect the flex/appliances that are plugged in. In the rest of the world the CU has to protect both.

This fusing of the plugs has another implication for CPC size as you can design the socket circuit to meet ADS and adiabatic without having to worry about the end appliance or use of extension leads. In the pre-RCD days you could never have an earth fault above a 13A fuse's capacity even on a high Zs end load, so a reduced size CPC is safe as it never has to clear overload-levels on the CU OCPD.

Polarity and earthing have been fundamental to the UK sockets from well before BS 1363 was introduced in 1947 and are often assumed “normal” here, but you can see cases around the world where polarity is not enforced (much of EU sockets can be inserted either way, some USA plugs, etc) or even places where an earth/CPC is not always present, or maybe not connected (seen in USA where it is allowed if RCD-protected on grand-fathered installations). In the UK generally it is only double-insulated appliances that have unpolarised plugs and sockets (e.g. some laptop power bricks, TVs, etc), never on the final wiring (maybe IT shaver supply excluded)

Because polarity and ADS are fundamental to safety here they must be proven on testing, more of this later.

[*] Actually BS 1363 (“Type G”) plugs and sockets are used elsewhere in the world, such as ROI, Malta, Malaysia, Singapore, etc, but that is a small portion of “everyone”.

 

[pc1966]

Example Installations

Here is an example of a UK 'pure' radial circuit for sockets:

And there is the 'pure' Ring Final Circuit:

Do you see how radically different they are? No?

Sarcasm aside, the only real difference is completing the loop back to the CU where the fixed wiring protection is, here shown as an RCBO in line with modern practice. But what does this mean for design? Or for testing? We will come back to testing later, for now we can look at what is needed from a circuit design point of view.

 

[pc1966]

Circuit design
The key aspects of any design are:

• We need to protect the cables from overload
• We need to meet the voltage drop limit of 5% at maximum expected load
• We need to meet ADS under earth-fault conditions

Points 2 and 3 are strongly related, but ADS with RCD protection is pretty easy even if many (like myself) ideally like to see it met on the OCPD side as well. It is the 1st point where it become more contentious due to the normal practice with the RFC of assuming tolerable load distribution so you can use 32A OCPD with 2.5mm cable, where that is usually only good for a bit over 20A, and the resulting small risk of overload from multiple spurs.

Using the typical common UK practice as set out in the on-site guide we have the following:

• RFC using 32A protection and 2.5mm cable, max length 106m
• Radial using 20A protection and 2.5mm cable, max length 42m

The eagle-eyed reader will note in the OSG that for the RFC the loads are assumed to be distributed, for the radial it is end-point. But as you can hit 20A by design/test on a single double socket, it is fair to use that in sizing the radial length. Of course you can use other combinations such as:

• RFC using 32A protection and 4mm cable, max length 171m
• Radial using 20A protection and 4mm cable, max length 69m
• Radial using 32A protection and 4mm cable (method C only), max length 43m
• Radial using 32A protection and 6mm cable, max length 63m (I personally would not try this as it is a pain to wrangle, even though many have terminals rated to accept 2*6mm).

What is apparent is the RFC makes the acceptable length 4 times greater for the same OCPD and cable size, basically due to the parallel paths that are normally also serving sockets in different areas. If your design requires sockets in a layout that is vaguely a loop, this is an excellent match.

 

[pc1966]

Testing of the Radial

A common complaint is that testing of the RFC is much more complicated than for the Radial. This is true, but due to the specified RFC test procedure also covering far more faults than the minimum needed for the radial. So what must you test in both cases? Basically:

• You prove polarity
• You establish ADS is met

Now the UK places very great importance on polarity (i.e. that L really is the line, and N really is earth-referenced) due to the long-established practice of only placing OCPD in the line conductor(s). If you swap L & N then you have no effective over-current protection and then a “N” to E fault is a serious fire risk. Also in pre-RCD days you lose any meaningful ADS that is supposed to protect against shock.

The most commonly accepted process for the radial socket is you link L to CPC at the CU end and then check the R1+R2 measurement at each of the sockets. By doing so a N-L swap will show up as an open circuit, and if the R1+R2 is low enough (i.e. below [OCPD Zs] - [supply Ze]) you know it should meet ADS time under a hard fault.

Now you can do exactly the same for the RFC and achieve exactly the same fault coverage, but that is not how it is specified since it becomes practical to perform a far greater level of fault coverage due to having both ends of the circuit conveniently accessible at the CU.

So what is missing from the radial's test? We have:

• Usually N is not tested. True, if the socket is dead with live tested, you know you have a fault But you don't actually have that good high-resolution DC resistance test to check the L and N are virtually identical, implying no poor high-resistance joint.
• If R1+R2 is below ADS it is not proof that L has no poor joints, nor that the CPC has no poor joints. Just that both together are currently good enough to trip the MCB.

We will return to this later.

 

[pc1966]

RFC testing

Testing the RFC is typically done in two ways. the first is to measure the three end-end resistance:

• r1 = (L1 - L2)
• r2 = (CPC1 - CPC2)
• rN = (N1 - N2)

Assuming the cable will have identical L & N conductors we expect r1 and rN to be identical DC resistance, at least to better than 0.05 ohms or so. If not there is something wrong (bad joint) or very odd in the cable route (unequal lengths in conduit?)

Finally we expect r2 to be in-ratio to r1 & rN with the cable used, so typically 2.5mm T&E has a 1.5mm CPC so we expect the resistance ratio to be 2.5mm/1.5mm = 1.67 if all is well. Too high and probably CPC has a poor joint, too low and you might be looking for parallel paths which, outside of metallic conduit, is very odd.

This test is easy to do during an ECIR as well, and again gives good fault coverage. What it does not do is prove polarity, which is even more fundamental to safety in the UK system.

Now you can do the same radial test of linking L & CPC to check polarity at each socket, but the folks who invented the RFC realised there is a far more cunning trick to be used here and that is the “figure of eight” approach where you link L1 with CPC2, and link L2 with CPC1. Why is this so clever?

Well now at every socket when you measure L-CPC resistance you should have an identical length of conductor between them and so every “R1 + R2” check this way ought to be identical. Anything even slightly higher is either a fault/poor socket or evidence of a spur with some additional cable inserted.

The only real test missing from the standard RFC procedure is to do the same figure-of-eight with the N to check for poor neutral contacts in any sockets, but the rN has already established good connections at least to each socket's terminals.

 

[pc1966]

Better radial testing?

Can you do better fault coverage with a radial circuit? Of course, it is just not as easy to do. The end-end values could be checked using a nulled wander-lead to the final socket, but that might be impractical/unsafe due to trip hazard. Alternatively you could do the following:

• Link L & CPC to measure R1+R2 (already done as standard)
• Link N & CPC to measure RN+R2
• Link L & N to measure R1+RN

You could do the 2nd pair at once (i.e. a bit faster) by linking L-N-CPC but that must not be done for the first test, otherwise you fail to prove polarity.

If R1+R2 and RN+R2 are nearly identical then you know the live conductors must be the same, so probably good. Then you can subtract (R1+RN)/2 = R1 = RN from either to get R2 and check the ratio as for the RFC.

Alas, the figure-of-eight test is probably impractical without two identical wader leads!

 

[pc1966]

An Infestation of Spurs

Both radial and RFC illustrated and discussed above are “pure” in the sense they have a single start and end point, with the RFC ending at the start. However, in practice you find circuits taken from points along the main route and these are known as spurs for the RFC and usually as branches for the radial. They may be convenient but that have some significant implications for safe design and for testing.

In the case of the RFC the fundamental issue is the common case of 2.5mm cable off a 32A OCPD, this is not providing adequate overload protection and so any spur must be limited to 20A max which is a double socket in practice. Yes, you could get 26A from that but it would be an odd case, though maybe possible in a kitchen with two 3kW appliances side by side.

Where it becomes a problem is when DIY or incompetent “electricians” add spurs off spurs without an understanding of the RFC assumption of moderate current-sharing between the two legs or of the OCPD to cable CCC aspect.

The same branching off from a radial is also possible but due to the end-fed arrangement it already has to be OCPD protected for a single run of the cable, typically 20A for 2.5mm. In this case overload is not an issue, nor would it be if the RFC was fitted with 20A OCPD as well.

The presence of spurs/branches also has implications for testing in both radial and RFC, as you cannot simply check the end-end values or final socket R1+R2 to prove, for example, CPC continuity to all intermediate sockets. Now it should be good practice to test every socket anyway, but we know that is often skimped upon.

OK, enough from me now and over to folks for (hopefully) a civilised and informed discussion.

 

[Dustydazzler]

Once a Ring is Infected with numerous spurs on spurs on spurs , it can no longer be called a Ring and must be called a Spider circuit which needs to added to the 19th Edition assuming Rings aren't outlawed before that

 

[timhoward]

Thanks for taking the time to write all that.

I hesitate to make this point but I think it is a valid consideration.
On many occasions I have found a high r1, rn or r2 on an RFC. Subsequent investigation finds the problem connection, and all is well.
The thing is, this has usually been the case for a very long time. The very fact it is a ring has in all likelihood minimised the consequences of this bad connection.

A similar loose connection on a radial is more likely to have noticeable effects if a heavy load if connected downstream of it. As an extreme example we fairly often see burnt or charred N connections on shower pull cords where the installer got fed up trying to tighten a connection and gave up.

Maybe a ring final circuit is more forgiving of human nature in general, whether it's the original installer making a mistake, or a DIY addition not going so well.

I'm well aware that a 2.5 sq mm T+E cable isn't rated for 32 amps if a ring gets severed but you might remember a thread a while back where I found an 8kw shower on a 2.5 sq mm supply. I was in such denial that this could possibly be ok that I convinced my eyes to believe it was a 4mm. The thing is, there were no signs of thermal overload anywhere, and it had been fitted and used regularly for 6 years.

I hope my point has come over correctly, I'm not condoning broken rings, high resistance connections, or undersized cables, and I fix them all. I'm just pointing out that in reality having a 2nd chance of getting back to the CU safely isn't a bad thing and even a broken ring (while not technically safe) usually comes to no harm in practise.

 

[pc1966]

Once a Ring is Infected with numerous spurs on spurs on spurs , it can no longer be called a Ring and must be called a Spider circuit which needs to added to the 19th Edition assuming Rings aren't outlawed before that

If you have a badly butchered RFC where properly fixing it is impractical then simply dropping the OCPD to 20A is a fix.

Yes, there might be an issue of total load but unless folks rely on heaters off sockets (instead of fixed electric, or gas/ASHP sort of system) it is unlikely. I have seen 20A RFC before off rewirable fuses, I guess due to high-ish Ze at the original design time, and they worked fine for decades.

 

[pc1966]

I hope my point has come over correctly, I'm not condoning broken rings, high resistance connections, or undersized cables, and I fix them all. I'm just pointing out that in reality having a 2nd chance of getting back to the CU safely isn't a bad thing and even a broken ring (while not technically safe) usually comes to no harm in practise.

Those are good points about fault tolerance. In these days of RCD everywhere it might be less of an aspect, but I have seen an RFC with open CPC (r2) and all sockets still had an acceptable CPC to meet ADS so remained safe.

On a radial with open CPC the first you would know (other than a sparky testing it) might be last you ever knew!

 

[Dustydazzler]

If you have a badly butchered RFC where properly fixing it is impractical then simply dropping the OCPD to 20A is a fix.

Yes, there might be an issue of total load but unless folks rely on heaters off sockets (instead of fixed electric, or gas/ASHP sort of system) it is unlikely. I have seen 20A RFC before off rewirable fuses, I guess due to high-ish Ze at the original design time, and they worked fine for decades.

Come across ALOT of old Council houses in my area which have 1 Ring for the entire houses sockets, which have been seriously infected now with multiple spurs on spurs etc.
De rating the OCPD to 20amps will often cause nuisance tripping when the house is just one Ring for every socket in the house.
I call these completely fooked Rings

 

[pc1966]

De rating the OCPD to 20amps will often cause nuisance tripping when the house is just one Ring for every socket in the house.

If it includes a modern kitchen then you have trouble. The fact the kitchen fitters did it badly is sadly par for the course

 

[telectrix]

all depends on the load. up to 27A ( less if derated of a ring. arting factors involved) a radial is king. personally, 'd rather have 2/3 radials i advanta