Because that is how the servers are powered.
5-10kW per rack is not unusual, and high-density racks can easily be double that - I will leave it as an exercise for the reader to work out what cable sizes would be needed for that much power at ELV DC.
What do the regs say about sockets fed from portable UPS being RCD Protected?
- Thread starter eveares
- Start date
5-10kW per rack is not unusual, and high-density racks can easily be double that - I will leave it as an exercise for the reader to work out what cable sizes would be needed for that much power at ELV DC.
- Joined
- 27 Jan 2008
- Messages
- 24,313
- Reaction score
- 2,777
- Location
- Llanfair Caereinion, Nr Welshpool
- Country
5 kW at 50 volt which is what telephones use is 100 amp but at 12 volt that is 416 amp and in the main even the mains UPS still work on 12 volt batteries, although unlikely at 5 kW and at 10 kW even 50 volts the 200 amp seems a bit on the high side. But it is unlikely that anything over 75 volt is used simply because of the rules at being classed as extra low voltage, so really does not matter if the unit supplies 50 volt DC or used 50 volt DC to produce 230 volt AC the same current is involved.
At 1.2 volts per cell a Ni/Iron battery needs 40 cells all bolted together with large nuts and banjo type links and the Lead acid is not any better with 24 cells at 2 volt each. The inverter needs multi-FET's to handle that current and the big problem with them is to get them to power share so the higher the voltage the better. The idea is not to have a power outage so having an UPS using 12 volt increases the risk of a FET failure so is not really a good idea.
We want a steady voltage even when the battery is becoming discharged so it will always need some switch mode power supply to turn the 50 volt to the 5 and 12 volts used by the server. But to transform up to 230 volt turn to AC, then turn back to DC and transform back down seems daft.
At 1.2 volts per cell a Ni/Iron battery needs 40 cells all bolted together with large nuts and banjo type links and the Lead acid is not any better with 24 cells at 2 volt each. The inverter needs multi-FET's to handle that current and the big problem with them is to get them to power share so the higher the voltage the better. The idea is not to have a power outage so having an UPS using 12 volt increases the risk of a FET failure so is not really a good idea.
We want a steady voltage even when the battery is becoming discharged so it will always need some switch mode power supply to turn the 50 volt to the 5 and 12 volts used by the server. But to transform up to 230 volt turn to AC, then turn back to DC and transform back down seems daft.
But not the same current to supply the rack of servers.
Not as daft as trying to supply an aisle of racks with 50-200kW at ELV.
- Joined
- 27 Jan 2008
- Messages
- 24,313
- Reaction score
- 2,777
- Location
- Llanfair Caereinion, Nr Welshpool
- Country
Are you trying to say that the computers in the racks run on 230 VAC? Most I am sure use a standard computer power supply:
Power Output Specifications
+3.3V Rail 22A (150W)
+5V Rail 16A (150W)
+12V1 Rail 17A (408W)
+12V2 Rail 17A (408W)
+5Vsb Rail 2A (10W)
-12V Rail 0.3A (3.6W)
Total Power Output 500W
So what we are doing is taking a battery say 48 volt, then with an inverter transforming it to 230 volt then with another inverter transforming in down to what the electronics use. Using a power supply with a 48 VDC input to get the outputs required directly rather than stepping up then down again would remove one of the electronic units and so also remove one unit which may fail.
If each unit with in the rack was 48 volt DC there would be no problem with floating power or any of the other considerations when using an UPS, it would be the same as any standard telephone exchange which also use rack systems and like the telephone exchange would be able to run without outside power for an extended time. Having stand by generators feeding 48 volt into the batteries direct missing out the low voltage completely would make so much more sense.
Yes today the telephone exchange has batteries in the rack rather than use a central battery bank, but again having servers designed to be battery powered from word go rather than try to use a unit not designed for battery back up and add the back up latter would be so much better.
I have a PC designed with battery back up, it is supplied with 19 volt, the genetic name is a laptop, although too big to really use on my lap, it is designed from the start to run on a DC supply. I can select a power supply, it can be a 100 volt to 250 volt AC/DC input or a 11 to 15 volt DC input, the latter clearly designed to connect to a 12 volt battery. It is designed from concept to be battery backed.
If servers need battery back up then they should be designed in the same way, not add an UPS afterwards as an after thought.
Power Output Specifications
+3.3V Rail 22A (150W)
+5V Rail 16A (150W)
+12V1 Rail 17A (408W)
+12V2 Rail 17A (408W)
+5Vsb Rail 2A (10W)
-12V Rail 0.3A (3.6W)
Total Power Output 500W
So what we are doing is taking a battery say 48 volt, then with an inverter transforming it to 230 volt then with another inverter transforming in down to what the electronics use. Using a power supply with a 48 VDC input to get the outputs required directly rather than stepping up then down again would remove one of the electronic units and so also remove one unit which may fail.
If each unit with in the rack was 48 volt DC there would be no problem with floating power or any of the other considerations when using an UPS, it would be the same as any standard telephone exchange which also use rack systems and like the telephone exchange would be able to run without outside power for an extended time. Having stand by generators feeding 48 volt into the batteries direct missing out the low voltage completely would make so much more sense.
Yes today the telephone exchange has batteries in the rack rather than use a central battery bank, but again having servers designed to be battery powered from word go rather than try to use a unit not designed for battery back up and add the back up latter would be so much better.
I have a PC designed with battery back up, it is supplied with 19 volt, the genetic name is a laptop, although too big to really use on my lap, it is designed from the start to run on a DC supply. I can select a power supply, it can be a 100 volt to 250 volt AC/DC input or a 11 to 15 volt DC input, the latter clearly designed to connect to a 12 volt battery. It is designed from concept to be battery backed.
If servers need battery back up then they should be designed in the same way, not add an UPS afterwards as an after thought.
For what is it worth it used to be ( and probably still is ) common in electronic telephone exchange to distribute the power as 48 volt DC to the ends of rack shelves. Convertors fed by 48 volt DC produce the various voltages required by the elctronic circuitry. The design of the convertors had to ensure that no electrical noise ( audiable frequencies ) was passed back to the DC supply as it provde the power to the loops to the telephones. The convertors were also synchronised by a central clocking system to ensure a smooth loading on the 48 volt supply.
Feeding mains to the racks was considered as a bad thing for both safety and mains hum on phone lines.
Not aware of that ( other than as memory back up )
Feeding mains to the racks was considered as a bad thing for both safety and mains hum on phone lines.
Not aware of that ( other than as memory back up )
No.
Most will, or something very similar - even with non-Intel based systems, memory is memory, PCI cards are PCI cards, disks are disks, etc.
It would.
Do the maths, Eric. 5/10/20kW @ 48V is 105/210/420A.
I remain unconvinced that needing to supply a rack using 70-185mm² cables instead of 4-10mm² makes much more sense.
And that's just to the rack - OK, you could use multiple PDUs with multiple power cords, and reduce the size, but what are you going to need under the floor to supply a row of racks with 50-100-200kW @ 48V?
The supply of 48 volts to the racks is normally by large bus bars. main bus bar from the battery room can be as large as 6 inch by 2 inch
Indeed.
7700mm²
7700mm²
Certainly was - I used to work at Baynards House, which is one of the main BT exchanges for the City. When the exchanges went electronic, the several floors full of equipment were replaced with a single room, the rest being converted into offices. I was there when they stripped out the old battery power systems and the busbars - BT must have made quite a lot of money selling the copper off!
A few random comments to the above ...
Yes, there are many different arrangements that are labelled "UPS" :
At the smaller end they are pretty well all "standby" systems where an inverter is started up and the load is transferred to it by a transfer switch. Some will keep the inverter running (on no load) and synchronised to the mains so as to allow faster switchover, but most don't.
Having an AC-DC converter, DC bus, and DC-AC inverter is generally called double conversion and gives much better isolation from mains "sh*t" like spikes etc - as well as zero transfer time since the load is never not already running from the inverter. The main reasons for not using double conversion are :
And in general, the higher the power, the higher the battery voltage - our 16KVA system at work uses nominal 120V (so 136V float) batteries.
And there is quite a bit of IT kit which can (either by selecting a DC variant, or adding additional module(s)) run off DC - usually 48V for commonality with telecoms supplies.
I have to admit, I have a couple of times wondered about the possible short circuit current available from the batteries on our system - 120V, cabinet full of low-impedance AGM batteries. Each string has a 60A HRC fuse, I forget what rating the main cabinet fuse is.
Yes, there are many different arrangements that are labelled "UPS" :
At the smaller end they are pretty well all "standby" systems where an inverter is started up and the load is transferred to it by a transfer switch. Some will keep the inverter running (on no load) and synchronised to the mains so as to allow faster switchover, but most don't.
Having an AC-DC converter, DC bus, and DC-AC inverter is generally called double conversion and gives much better isolation from mains "sh*t" like spikes etc - as well as zero transfer time since the load is never not already running from the inverter. The main reasons for not using double conversion are :
- You need full load, unlimited time, rated converters - whereas a standby system only need an inverter rated for as long as the batteries last, and the battery charger can be much lower rated.
- You are doing two conversion which both involve losses - so you are permanently losing a few % of your supply in heating up the room.
And in general, the higher the power, the higher the battery voltage - our 16KVA system at work uses nominal 120V (so 136V float) batteries.
And there is quite a bit of IT kit which can (either by selecting a DC variant, or adding additional module(s)) run off DC - usually 48V for commonality with telecoms supplies.
I have to admit, I have a couple of times wondered about the possible short circuit current available from the batteries on our system - 120V, cabinet full of low-impedance AGM batteries. Each string has a 60A HRC fuse, I forget what rating the main cabinet fuse is.
- Joined
- 27 Jan 2008
- Messages
- 24,313
- Reaction score
- 2,777
- Location
- Llanfair Caereinion, Nr Welshpool
- Country
So it does seem larger systems use DC which makes sense easy to have three or four supplies all feeding a common DC rail.
However the question about earthing with the smaller systems is still valid. Single socket outlet from an UPS is OK. But once one has two or more outlet sockets, unless feed from separate output windings on the transformer we should in theroy be earthing the output neutral.
But how can any manufacturer at reasonable cost earth the neutral making it a true neutral without the risk of not removing that earth when running on direct mains? To my mind it can't be done.
When we look at temporary supplies we often see non compliance with regulations. The simple yellow brick 110 volt transformer is a typical example where the only overload is on the input. We all know the output should be fused, but for any tradesman to have both 230 volt and 110 volt tools is not really an option and so when not on a large site with 110 volt supplies from a central transformer the yellow brick fills the gap.
The same applies to the small UPS supply, used once in a blue moon it is unlikely to be a problem with a floating IT supply. And as the electrician there is little or nothing we can do, it is down the the manufacturer rather than the installer to ensure the device is safe to use.
The question arises however what is the point where we should say enough is enough this is the limit? Taking a pen home from work is likely considered as OK, taking a computer home is likely considered as theft. Both are really theft but no one worries about the pen.
So server behind a cage with UPS is OK, no unauthorised assess. A small UPS under each desk is OK, unlikely to get two items go faulty at the same time. But a UPS feeding a whole office is likely going OTT. In an office with 100 items on an UPS the chances of two items going faulty within the time that UPS power is being used is reaching a point where a risk assessment may show it as an unacceptable risk, specially since using individual UPS under each desk is not any more expensive.
If we consider a personal work station, likely PC, Monitor and maybe an external hard drive needs an UPS supply. Printers do not really need a UPS supply. So for a personal work station an UPS with 4 outputs is ample. There is really no case for pairing work stations, so there is really no case for using a UPS with more than 4 outputs. So doing a risk assessment I would consider it to be a low enough risk to have 4 items on an UPS with a IT output when running on battery back-up. With an 8 output UPS personally I would say it's going OTT.
So I would be happy to pass a 4 output UPS when PAT testing, with an 8 output I would only pass it for use in places with restricted access. What does everyone else think?
However the question about earthing with the smaller systems is still valid. Single socket outlet from an UPS is OK. But once one has two or more outlet sockets, unless feed from separate output windings on the transformer we should in theroy be earthing the output neutral.
But how can any manufacturer at reasonable cost earth the neutral making it a true neutral without the risk of not removing that earth when running on direct mains? To my mind it can't be done.
When we look at temporary supplies we often see non compliance with regulations. The simple yellow brick 110 volt transformer is a typical example where the only overload is on the input. We all know the output should be fused, but for any tradesman to have both 230 volt and 110 volt tools is not really an option and so when not on a large site with 110 volt supplies from a central transformer the yellow brick fills the gap.
The same applies to the small UPS supply, used once in a blue moon it is unlikely to be a problem with a floating IT supply. And as the electrician there is little or nothing we can do, it is down the the manufacturer rather than the installer to ensure the device is safe to use.
The question arises however what is the point where we should say enough is enough this is the limit? Taking a pen home from work is likely considered as OK, taking a computer home is likely considered as theft. Both are really theft but no one worries about the pen.
So server behind a cage with UPS is OK, no unauthorised assess. A small UPS under each desk is OK, unlikely to get two items go faulty at the same time. But a UPS feeding a whole office is likely going OTT. In an office with 100 items on an UPS the chances of two items going faulty within the time that UPS power is being used is reaching a point where a risk assessment may show it as an unacceptable risk, specially since using individual UPS under each desk is not any more expensive.
If we consider a personal work station, likely PC, Monitor and maybe an external hard drive needs an UPS supply. Printers do not really need a UPS supply. So for a personal work station an UPS with 4 outputs is ample. There is really no case for pairing work stations, so there is really no case for using a UPS with more than 4 outputs. So doing a risk assessment I would consider it to be a low enough risk to have 4 items on an UPS with a IT output when running on battery back-up. With an 8 output UPS personally I would say it's going OTT.
So I would be happy to pass a 4 output UPS when PAT testing, with an 8 output I would only pass it for use in places with restricted access. What does everyone else think?