This discussion is locked.
You cannot post a reply to this discussion. If you have a question start a new discussion

RCD Protection at Data Cabinets

I have a Client that will not provide a Risk Assessment to omit an RCD Commando Socket for a Data Cabinet. He wishes to employ a local isolator for the Cabinet, however he is concerned that the PDU Strip that arrived on site from the factory and pre-fitted within the data cabinet would still need an RCD. It is my opinion that the Factory built Data Rack is classed as a piece of equipment and as such does not form part of the Contractor's installation. If the Rack had arrived on site without the PDU strip fitted and it was installed by the Contractor then this would form part of the installation and therefore would require an RCD. Any advice/assistance would be appreciated.
  • Given BS EN 50310 (which is referred to from BS 7671), at least some of the rack power distribution is covered by BS 7671, as to the latest version of BS EN 50310 it requires local bonding to the rack earth bar, and the rack earth bar bonding to the room earthing etc.


    This does indeed ask the question whether RCD protection is required for the socket-outlets if no risk assessment is provided.


    The only other option open to you is to use rack power distribution units that employ couplers (to BS EN 60320) rather than socket-outlets to BS 1363 or BS EN 60309 - BS 7671 does not (at present) require RCD protection for couplers. So provided the power distribution strips comply with relevant standards ...
  • This does bring out a point from BS7671, which is the blanket requirement for RCDs for sockets up to 32A. In an installation such as this with supplementary bonding to earth, the RCD has exactly zero use! There is no shock risk unless the equipment is opened and live terminals accessed, whatever earth fault occurs. This point needs revision, because it simply doesn't make any kind of safety sense. From your point of view I suggest fitting 63A BS4343 connectors to the installation. If they choose an RA, it is fairly straightforward, the environment is controlled and nothing else may be plugged into the socket, and the supplementary bonding must be in place and tested (notice) before the equipment is energised.

  • davezawadi:

    and the supplementary bonding must be in place and tested (notice) before the equipment is energised.




    I feel the need to reinforce the fact that, for the assumption that risks are controlled to be true, ALL of the supplementary bonding must be in place, including the connection between the rack earth bar and the PE of the power distribution strips, though. Whilst this is now clearly stated in the latest version of BS EN 50310, I can't tell you how many racks I've seen, and continue to see see without these connections made.


    It's also worth noting the following:


    • if you go back to BS EN 61140, there should be no reason RCDs can't be used for individual socket-outlets (except the discussion we've had in the Forums already about the 18th Edition that SRCD standards aren't listed in the relevant Regs in Chapter 5 of BS 7671 ?)

       

    • the design decision to connect a "rack" using plug and socket-outlet rather than a fixed wiring connection point is really a preference. The often-used statement that "it makes it easier to install and remove racks as we need them" doesn't hold water, as to comply with BS EN 50310, you need to make fixed wiring connections for the CBN bonding to each rack, and execute relevant electrical tests stated in that standard.

  • Well,  except for the potentially significant convenience of not having to turn the power off and verify exposed terminals are dead before unbolting things - as with signal cables, bonding can be removed and re-fitted, as soon as the jumbo mains plug is removed from the supply...  Yes you can have local isolation switches one per rack, but these in the heat of battle such things can be another source of accidentally switching off the wrong thing.

  • Well,  except for the potentially significant convenience of not having to turn the power off and verify exposed terminals are dead before unbolting things - as with signal cables, bonding can be removed and re-fitted, as soon as the jumbo mains plug is removed from the supply






    As long as you're not importing protective conductor currents from elsewhere.




    Yes you can have local isolation switches one per rack, but these in the heat of battle such things can be another source of accidentally switching off the wrong thing.






    Same with plug and socket-outlet. They're all just isolators. And many racks will have two feeds (or more) - multiple points of isolation. In terms of accidentally switching off the wrong thing in the heat of the moment, let's also consider Regulation 12 of the Electricity at Work Regulations 1989 which definitely requires "... where appropriate, methods of identifying circuits ...": if it's operationally so important not to "flip the wrong switches", make sure they are properly identified, because this is required for safety in any case.



  • This does bring out a point from BS7671, which is the blanket requirement for RCDs for sockets up to 32A. In an installation such as this with supplementary bonding to earth, the RCD has exactly zero use!



    But supplementary bonding won't protect against instances of direct contact - e.g. from a damaged flex - where a 30mA RCD will provide some benefit. I've certainly seen flexes trapped in 19" rack doors and pulled around sharp edges of some of the metalwork before now. Certainly they're the situations it's easy to avoid with a bit of care - a policy of properly securing flexes and only allowing access to (authorized) persons with a bit of nous would seem to do - but we can't really assume that without a risk assessment to say that's the case. A lot of IT types, which very technical in some ways, really don't appreciate the hazards of mains electricity.


    Same with plug and socket-outlet. They're all just isolators.



    Sorry, but I've got to disagree with that. Plugs & sockets mean that IT technicians and change the power supply arrangements safely. Bare in mind that when something goes wrong with live (i.e. customer facing) systems in the IT world there is huge pressure on technicians to get things working as quickly as possible - calling in an electrician who might not turn up for a few hours simply isn't an option. Being able to unplug a rack from one supply and plug it into another nearby supply has saved things on my watch at least once.


      - Andy.

  • AJJewsbury:


     


    Same with plug and socket-outlet. They're all just isolators.



    Sorry, but I've got to disagree with that. Plugs & sockets mean that IT technicians and change the power supply arrangements safely. Bare in mind that when something goes wrong with live (i.e. customer facing) systems in the IT world there is huge pressure on technicians to get things working as quickly as possible - calling in an electrician who might not turn up for a few hours simply isn't an option. Being able to unplug a rack from one supply and plug it into another nearby supply has saved things on my watch at least once.


      - Andy.

     




    Well ... the original discussion point was that it was easy to get "mixed up" with an isolator, and that is what I was answering.


    I definitely get the point about quickly connecting to another supply, but of course that may mean making a decision whether it's safe to do so.


    It is still the case that making an assembly that doesn't really meet BS EN 61140 (in that the protective conductor current is too high, evidenced by the fact you can't use a 30 mA RCD) and then trying to treat it as a 32 A (or less) pluggable appliance, doesn't quite stack up. Basically, for pluggable equipment, the limit in BS EN 61140 for protective conductor current with rating > 20 A is 10 mA. To exceed this limit, the equipment should be permanently connected, AND have a reinforced protective conductor, AND be limited to a protective conductor current of 5 % of the phase current.


     


  • Beard Weird:



    This would have been an ideal topic to be forwarded to the IT section of the IET forum!


     




    It would certainly be an interesting discussion.


    edit: In fact, I've just posted a link to this Thread from the Data Centre Engineering Community discussions.

  • I think that Andy has also made an error worrying about direct contact with damaged cables etc by anyone. Whilst that could happen given enough various circumstances, realistically it is very very unlikely. I would point out that a single RCD for a rack of servers (which might take 100A of mains, probably 3x32A supply) would be foolish, the idea of fitting one each is worse! We have an electrical system which depends on proper earthing of exposed conductive parts, and automatic disconnection of supply by CPDs, yet this regulation is determined to fit even more unnecessary RCDs! The scope for RCD protection is way too wide, it should only be for sockets where portable appliances are likely to be used by unskilled persons, and the servers in a rack should be out of scope for portable appliances separately, as you need tools to make them portable.


    There is a serious problem with the standards making process, and that is that interactions between various ones from differing sources lead to foolish and unnecessary outcomes. The most obvious one is electric cars (chargers) not being class 2 because someone didn't see that they could not safely be earthed in the external environment, which we all know and have a lot of regulations to reduce risks to presumably acceptable levels. However street furniture using PME is considered entirely satisfactory and does not cause a significant number of accidents. We need to be consistent on the basis of real risk only, not on "what if" scenarios.

  • I think that Andy has also made an error worrying about direct contact with damaged cables etc by anyone. Whilst that could happen given enough various circumstances, realistically it is very very unlikely. I would point out that a single RCD for a rack of servers (which might take 100A of mains, probably 3x32A supply) would be foolish, the idea of fitting one each is worse! We have an electrical system which depends on proper earthing of exposed conductive parts, and automatic disconnection of supply by CPDs, yet this regulation is determined to fit even more unnecessary RCDs! The scope for RCD protection is way too wide, it should only be for sockets where portable appliances are likely to be used by unskilled persons, and the servers in a rack should be out of scope for portable appliances separately, as you need tools to make them portable



    I agree RCDs aren't the solution - but I don't agree that the risk can be just ignored either. I've seen too many IT racks where things had obviously been put together in a very hap-hazard way, with everything from extra sockets provided by dangling supermarket 4-way extension leads, leads trapped in doors or slide rails, to mains leads with core insulation exposed. Pretty much all the horrors you see in some domestics. Also I don't think they're really in the usually-not-handled-when-live class of fixed appliances either - so much is hot swappable these days (usually PSU and fan modules are - some even allow hot addition of processors and memory) - and the idea of switching off before sliding a server out of the rack, or shoving an arm through the tangle of cables at the back to pull another lead through, just wouldn't be a consideration. If anything a restrictive conductive location might be a closer approximation. Like I said, I agree that 30mA RCDs aren't the answer, but I do see the risk assessment as a means of chivvying management into making sure their IT staff actually have a bit of basic electrical safety knowledge and implement some basic good practices.


      - Andy.