18

We've got 2 UPSes in our server room. One's 3 phase, and has a 3phase supply, yet the APC network stats thing only says

Incoming Voltage: 230.1 Volts

The other is a single phase one with no network-type monitoring.

What I'd really like to know, and can't seem to find out anywhere.. What does a UPS do with the 3 separate phases? Does it use both of them to charge the batteries, then provide a single sinewave output? This would be ideal, because then we'd be able to effectively use the UPS to convert 3 separate phases (which must remain so, else you get bangy power supplies and the magic smoke escapes).

For reference.. the 3-phase one is a "Smart-UPS RT 8000 XL".

Question the First: Why does the voltage on the 3-phase UPS' web interface say 230, not 415?

Question the Second: What on earth does the UPS do with 3 phases of power?

Yes, I've looked on APC's website, but couldn't find a definitive answer.

EEAA
  • 108,414
  • 18
  • 172
  • 242
Tom O'Connor
  • 27,440
  • 10
  • 72
  • 148

3 Answers3

13

A 3Ø UPS will have 3 phases coming in and 3 phases going out. You then put a 3Ø load onto the UPS. Typically that load will be a distribution panel.

As Chris S said, the reason you see 230V is that's the leg-to-leg voltage. Do note: you have three leg-to-leg circuits instead of one.

It's a common practice (at least it is here in .ca) to power a 3Ø distribution panel with the 3Ø UPS, then run individual single-phase or three-phase circuits from it. When powering single-phase circuits from your 3Ø distribution panel you do have to keep an eye on the amperage being used by each leg. Unbalanced load puts a large stress on the transformers.

Some of your datacenter loads are large enough to have 3-phase incoming, such as:

  • air conditioning system (large motors are well-suited to 3Ø)
  • high-end storage arrays (DS8800, XIV, etc)
  • large mainframe systems (a Big IBM Power server)
  • Blades (IBM and HP both have blade systems that take 3Ø)

Your smaller loads will run just fine on a single phase (two legs from the 3Ø), just make sure it's balanced. Your electricians should help. example setup


In your particular case the UPS can run on either 1Ø or 3Ø input but provides no 3Ø output.

What's going on behind the scenes is probably that some of those outputs are going to be the equivalent of L1-L2, some will be L2-L3 and some will be L3-L1.

The reason for this is that if you take the 3Ø input and process it entirely into a single waveform, that prevents a bypass from being used (as you no longer have any match between input and output power).

MikeyB
  • 38,725
  • 10
  • 102
  • 186
11

All UPSes work in this basic way:

Incoming Power       Bypass      (Optional)         Output
========+===========[SWITCH]=======[PFC]=========+========
        |                                        |
        \--[RECTIFIER]--[BATTERIES]--[INVERTER]--/

So the output will almost always be the same form of power as the input, so that the UPS can operate in Bypass when grid power is available. The only notable excption to this are certain DC UPSs which are crazy expensive (considering what they do) and very rare.

As for the 230v question; the difference in voltage between any two legs of 3Ø in the US is 230v or 208v (depends on your service). The difference from any leg to ground is the 120v that you're familiar with.

Side note: don't try to separate the legs of 3Ø into individual cuircuits. This will introduce a DC component to the UPS's internal tranformers (causing them to fail either soon or spectacularly). The same should be said about power from the grid, though 1. There's enough going on that it usually sort of balances 2. Your use/abuse is so small compared to the grid that the distortions would hardly be notices (obviously not the case with a UPS you're running at 50%+ load).

MikeyB
  • 38,725
  • 10
  • 102
  • 186
Chris S
  • 77,337
  • 11
  • 120
  • 212
  • We've got 80 odd workstations, and are about to get 20-odd dell M610e blade servers. We've got quite a load problem and can't do it without splitting the 3 phase into seperate circuits to neutral. We're in the UK, if that makes a difference. – Tom O'Connor Jan 16 '12 at 17:36
  • Dell's blade chassis doesn't have 3-phase power supply options?? Really? Also, the two of those UPSes aren't going to be enough for 80 workstations and 20 M610e (estimating 200W each on average)... You're probably between a rock and a hard place on this one. – Chris S Jan 16 '12 at 18:14
  • 1
    You don't want to split out into separate circuits to neutral; you want to split out into separate circuits as follows: (L1-L2, L2-L3, L3-L1) – MikeyB Jan 16 '12 at 18:32
  • Screw the workstations. We don't want them on the UPS. I was merely pointing out that we've got a lot of gear. – Tom O'Connor Jan 16 '12 at 19:30
  • It doesn't especially matter how you split the legs if you're going to do that. The key is that uneven loads ruin transformers. Splitting them into 3 230v 2Ø legs means that each of the three combinations splits it's load against 2 legs. So if you do have say 10A, 10A, 12A load; then one leg will by 1A off rather than 2A off. It's really a better idea to match your USPes to the actual load, 3Ø for 3Ø loads, 1Ø for 1Ø loads... but yeah. – Chris S Jan 16 '12 at 19:33
  • @Tom Under "normal" circumstances if your blade chasses have n/3 power supplies then you can wire them up as MikeyB points out and should have minimal problems. I'm really surprised Dell doesn't have 3Ø power supplies, but maybe I shouldn't be. – Chris S Jan 16 '12 at 19:34
  • @ChrisS Oh, you mean match the loads on the UPS's output side? – Tom O'Connor Jan 16 '12 at 20:52
  • @TomO'Connor Yep. Normally load balancing like that is very hard to do manually, which is why I strongly suggest against it, but it does work. – Chris S Jan 16 '12 at 21:11
5

I'm guessing MikeyB's answer applies to fairly high end setups.

Where I was last working, we had a 10KVA UPS installed which had inputs from each of the three phases, but the outputs were all single phase only.

The unit was an online UPS - i.e. all three phases were continually charging the batteries, then the battery power would be continually be converted to single phase AC.

I had hoped that if one phase went down the unit would carry on working on the remaining two phases but the installation engineer told me that if any of the phases had an issue the unit would run from battery.

For the UPS mentioned - Smart-UPS RT 8000 XL - the outputs are IEC C13 (kettle lead), IEC C19, and one three terminal output - these cannot be 3 phase outputs as they do not have enough conductors.

Robin Gill
  • 2,503
  • 13
  • 13
  • I wonder what the internal topology for that UPS is. Incoming Power - 3Phase Ups - Distribution Unit - 3x 120V out, or Incoming Power - Distribution Unit (3x 120V) - 3x 120V Ups - 3x 120V out? – Dan Is Fiddling By Firelight Jan 16 '12 at 19:11
  • Ah, in this case the various output plugs are all on different phases. Clarified my answer. – MikeyB Jan 16 '12 at 19:14
  • But were the outputs all on different phases? Is there a such a thing that could take 3 phase input, downcovert and present one single phase, with 3x the current handling capability of a single phase? – Tom O'Connor Jan 16 '12 at 19:29
  • Tom, this design is buildable, but would find few customers. A good UPS must "bypass" -- switch the load smoothly between UPS power and street power. (Handy when batteries die, inverters fail, or maintenance must happen.) Can you think of a way to switch a single phase load smoothly to a three-phase source, preferably with a glitch smaller than, say, 1/4 cycle of the AC waveform? – Paul Feb 14 '12 at 16:49