your preference for large number of computers, 120V or 240V

No question, get 240v wired L/L. That eliminates noise on the neutral and harmonic problems. Mainframe computer rooms didn't even bring a neutral into the panels back in the big iron days. Everything was

208/240.. Virtually all PC system units will switch to 240 and you can order 240v monitors and other equipment.

On Sun, 09 Nov 2008 22:28:58 -0500 snipped-for-privacy@aol.com wrote: | On 10 Nov 2008 02:18:03 GMT, snipped-for-privacy@ipal.net wrote: | |>If you are building a data center for a large number of computers (lots of PCs |>and server sized PCs, with all the network infrastructure), and this involved |>a large amount of power, and this were in a location where the standard user |>line-to-neutral voltage is 120V, but you could get a line-to-neutral voltage |>of 240V by special order, would you choose the latter to reduce wiring costs |>and I2R losses? If the size matters, where would the point of indecision be? | | No question, get 240v wired L/L. That eliminates noise on the neutral | and harmonic problems. Mainframe computer rooms didn't even bring a | neutral into the panels back in the big iron days. Everything was | 208/240.. | Virtually all PC system units will switch to 240 and you can order | 240v monitors and other equipment.

But not 240V wired L-N (as in 416Y/240) ?

Many mainframes I worked on back in the day used delta wired motor-generators and produced 400 Hz (I never found out the voltage at this point) for the CPU power systems.

For in home use, where the options are 120 L-N or 240 L-L, I'd like to go with the latter. Proper surge protection and UPS systems are hard to find for that configuration at home usage scale.

For a large business data center, likely to be powered by its own three phase service, choices would be 208Y/120, 416Y/240, and 480Y/277. The 416Y/240 can be made by having three 240V pole pigs, and 416Y/240 is an available option in pad mount transformers I have seen. The utility itself may pose complications in getting that. But once 416Y/240 is there, I'd then go with equipment that would be intended for places like Europe and Australia. All the UPSes I have seen for that voltage configuration can deal with 60 Hz. Surge protection shouldn't care about the frequency.

If I wanted 240V wired L-L, that might be possible with service metered at distribution (MV) voltage. Then 3 separate 120/240 transformers and "6 phase" wiring. But 416Y/240 still beats 3*120/240 on a large scale, with regard to minimizing conductor size and I2R loss. The odd 240Y/139 wouldn't do much better.

One possible option at large scale is for continuous dual conversion UPSes that power a small group of racks to work directly from 480 or 480 delta, and produce 120/240 directly, without any bypass option. Then 480Y/277 is usable.

Every data center I've designed in the past used 208Y/120 and everything was either running on 120 L-N or 208 L-L or 208 D. I focused more on grounding and protection than on power efficiency back then. I'm exploring what I might consider in these days of green data centers and high Cu prices.

| Virtually all PC system units will switch to 240 and you can order | 240v monitors and other equipment.

Oh, and my latest LCD monitor, and even my HDTV, support 100-240V 50/60Hz. Almost all my external device wall warts do, as well (but I'd have to be powering them through NEMA 5-15R outlets, which I'm not comfortable with for reasons of accidentally plugging in something that loses smoke on 240). I can get IEC power cables with NEMA 6-15P or Schuko plugs, for the computers and monitors. I might consider a larger 12VDC PSU for the external drives at home.

Why would you want to introduce the noise and harmonic problems you get with a neutral load?

The voltage on IBM systems was 208 on the 400 hz supply

They are availab;e in the commercial market but not normally in the residential market

The data center usually runs off the normal building service but they did bring IG into the data center panels, no neutral. I doubt they even use 3 phase in any but the very largest data centers these days.

These days data centers are not particularly big power hogs. The Ramac racks were the biggest load (A buttload of 3.5" hard drives in a rack) Now that multi T-byte drives are in the marketplace, they don't need that many drives. I pretty good size "data center" will go in a closet. That is one reason I got out of the business. You don't have to be an engineer to design a room with two or three racks in it and you don't have to be very skilled to fix a machine that has a blinking red light on the bad card.

You should be ordering NEMA 6-15 cords for them.

Not so fast there!

If I ruled that part of the world where you are wiring up a new computer center, I would have a mix of 120 and 240 outlets and/or 120/208 2 wire.

You don't want to shoot yourself in the foot and restrict yourself to only

240 volt displays, etc.

No matter how sophiscated an operation is, there isn't an harm in arranging things so that you can send someone down to Best Buy or Radio Shack for a display or some gadget that requires a wall wart and find out that the only stuff you can get RIGHT NOW is 120 volts.

Yeah, the higher voltage has many advantaes as much of the equipment as practical would run from the higher voltage.

Also, one would hope that if there is a "large number" of computers, there is provisions for key services to be maintained during a power failure. In that case, it might make sense for most of the computers to run directly off of a 24 or 48 volt battery. With a 3 phase battery charger you have have pretty good efficiecy and just survice the grid going down while you leisurely start your stnadby geneator.

Don't take this too seriously John. This is a Phil question and the thread will digress into absurdity soon enough. I agree, these days "computer rooms" are just a rack of PCs as often as not. PC may not be exactly correct but to the naked eye the RISC machines look like PCs.

For example, the large IBM Power 575 Hydro-Cluster installed at NCAR back in April uses 480V to each frame. There are 4064 Power-6 CPUs in 11 frames with a mix of air and water cooling employed. The total cluster uses about 700 KW of power.

What I'd do is run 120/208V to the data center, however I'd try to configure as much as possible to run off 208V, wired line to line, with everything balanced equally between each of the 3 possible pairs of phases. Lots of 6-15P cords and 6-15R or 6-20R recepticles. Why? Most computer power supplies will autoswitch between anything from

100-250V, but the real reason is they generate harmonics which tends to overload the neutral when large numbers of them run off 3 phase L-N (120V). If they're wired line-line, there's no neutral current, plus the power spike (which when wired L-N happens at the peak voltage) becomes two smaller power spikes on either side of the waveform peak (the peak between the phases).

There will be plenty of 120V power available, of course, for devices which do not have 120/240V autosense, things with wall warts, and whatever widget someone may need in the future that runs off "normal" power. Any of these with a switching PS will generate neutral harmonics, but (hopefully) not enough to overload, plus their line peak will happen at a different point of the line waveform than the big load.

On Mon, 10 Nov 2008 22:48:27 -0500 snipped-for-privacy@aol.com wrote: | On 11 Nov 2008 02:01:39 GMT, snipped-for-privacy@ipal.net wrote: | |>On Sun, 09 Nov 2008 22:28:58 -0500 snipped-for-privacy@aol.com wrote: |>| On 10 Nov 2008 02:18:03 GMT, snipped-for-privacy@ipal.net wrote: |>| |>|>If you are building a data center for a large number of computers (lots of PCs |>|>and server sized PCs, with all the network infrastructure), and this involved |>|>a large amount of power, and this were in a location where the standard user |>|>line-to-neutral voltage is 120V, but you could get a line-to-neutral voltage |>|>of 240V by special order, would you choose the latter to reduce wiring costs |>|>and I2R losses? If the size matters, where would the point of indecision be? |>| |>| No question, get 240v wired L/L. That eliminates noise on the neutral |>| and harmonic problems. Mainframe computer rooms didn't even bring a |>| neutral into the panels back in the big iron days. Everything was |>| 208/240.. |>| Virtually all PC system units will switch to 240 and you can order |>| 240v monitors and other equipment. |>

|>But not 240V wired L-N (as in 416Y/240) ? |>

| Why would you want to introduce the noise and harmonic problems you | get with a neutral load?

These are not really eliminated by L-L loads. All L-N does is concentrate the issues when you have a shared neutral. You need to have a panels that handle large neutral current, and wiring upstream from there accordingly. Branch circuits would have separate, non-shared, neutrals.

|>Many mainframes I worked on back in the day used delta wired motor-generators |>and produced 400 Hz (I never found out the voltage at this point) for the CPU |>power systems. |>

| The voltage on IBM systems was 208 on the 400 hz supply

Nice to know. Since it was a derived system, they could have chosen to make it be whatever they wanted to design their power supplies for.

|>For in home use, where the options are 120 L-N or 240 L-L, I'd like to go with |>the latter. Proper surge protection and UPS systems are hard to find for that |>configuration at home usage scale. | | They are availab;e in the commercial market but not normally in the | residential market

And probably for 208 instead of 240.

| These days data centers are not particularly big power hogs. The Ramac | racks were the biggest load (A buttload of 3.5" hard drives in a rack) | Now that multi T-byte drives are in the marketplace, they don't need | that many drives. I pretty good size "data center" will go in a | closet. That is one reason I got out of the business. You don't have | to be an engineer to design a room with two or three racks in it and | you don't have to be very skilled to fix a machine that has a blinking | red light on the bad card.

The largest one I worked on was 3600 machines back then when I left in 1997, and reportedly has broken the 10000 machine barrier. I'm sure it mostly blade type technology these days, so it could well be a lot less power than before. But it's no closet operation. Other places I worked on included ISPs with as many as 20 rows of racks (lots of routers, dial-in boxes, servers, etc).

When I say "big", I don't mean a closet. Most businesses can put their "data center" in a closet if they are pressed for space (and many more even if not). I'd rather spread things out if I can for the smaller ones, especially if *I* am the one that has to get into the machines to do work :-)

On Mon, 10 Nov 2008 22:49:58 -0500 snipped-for-privacy@aol.com wrote: | On 11 Nov 2008 02:06:54 GMT, snipped-for-privacy@ipal.net wrote: | |>Oh, and my latest LCD monitor, and even my HDTV, support 100-240V 50/60Hz. |>Almost all my external device wall warts do, as well (but I'd have to be |>powering them through NEMA 5-15R outlets, which I'm not comfortable with |>for reasons of accidentally plugging in something that loses smoke on 240). |>I can get IEC power cables with NEMA 6-15P or Schuko plugs, for the computers |>and monitors. |>

|>-- | | You should be ordering NEMA 6-15 cords for them.

You mean a cord with a 6-15P on one end and a 5-15R on the other?

I have about 10 wall warts in use right now that have NEMA 1-15P integrated into the PSU box itself (the true wall wart). Most of them are the "side" type where come off sideways, so a traditional power strip can handle them in all the outlets. Now days many power strips turned the outlets 90 degrees so the conflict is back. But I found some nice metal-housing power strips that have conventional duplex devices inside. These are oriented in the traditional way, so my "side plug" wall warts fit fine. I would need to change the surge protection in them to deal with 240V.

Alternatively, I could acquire the equivalent wall warts with Schuko plugs.

Or I could go with a big common 12VDC supply system instead of all those wall warts. I just don't know what risks that might have with respect to surges on the DC when plugging and unplugging devices into the DC bus. OTOH, DC does make surge filtering a little easier: a big capacitor with the appropriately sized resistors.

On Tue, 11 Nov 2008 01:59:23 -0500 snipped-for-privacy@aol.com wrote: | On Tue, 11 Nov 2008 00:02:15 -0500, "John Gilmer" | wrote: | |>

|>

|>>

|>> No question, get 240v wired L/L. That eliminates noise on the neutral |>> and harmonic problems. Mainframe computer rooms didn't even bring a |>> neutral into the panels back in the big iron days. Everything was |>> 208/240.. |>> Virtually all PC system units will switch to 240 and you can order |>> 240v monitors and other equipment. |>

|>Not so fast there! |>

|>If I ruled that part of the world where you are wiring up a new computer |>center, I would have a mix of 120 and 240 outlets and/or 120/208 2 wire. |>

|>You don't want to shoot yourself in the foot and restrict yourself to only |>240 volt displays, etc. |>

|>No matter how sophiscated an operation is, there isn't an harm in arranging |>things so that you can send someone down to Best Buy or Radio Shack for a |>display or some gadget that requires a wall wart and find out that the only |>stuff you can get RIGHT NOW is 120 volts. |>

|>Yeah, the higher voltage has many advantaes as much of the equipment as |>practical would run from the higher voltage. |>

|>Also, one would hope that if there is a "large number" of computers, there |>is provisions for key services to be maintained during a power failure. In |>that case, it might make sense for most of the computers to run directly off |>of a 24 or 48 volt battery. With a 3 phase battery charger you have have |>pretty good efficiecy and just survice the grid going down while you |>leisurely start your stnadby geneator. |>

|>

| | Don't take this too seriously John. This is a Phil question and the | thread will digress into absurdity soon enough.

I wasn't going to eliminate 120V outlets in my setup at home, nor would I do so in any data center I put together. OTOH, I would not be comfortable with setting up office desktop computers to run on 208V or 240V. As you know, no matter how much you idiot-proof things, they just keep coming up with better idiots. A 240V 6-15 outlet anywhere NOT near a window air-conditioner is just inviting someone to figure out how to plug something into it, that shouldn't be plugged into it.

Phil: I want the 6-20R for the A/C over here. Electrician: But this is nowhere near the window.

On Wed, 12 Nov 2008 23:54:16 +0000 (UTC) Michael Moroney wrote: | snipped-for-privacy@ipal.net writes: | |>If you are building a data center for a large number of computers (lots of PCs |>and server sized PCs, with all the network infrastructure), and this involved |>a large amount of power, and this were in a location where the standard user |>line-to-neutral voltage is 120V, but you could get a line-to-neutral voltage |>of 240V by special order, would you choose the latter to reduce wiring costs |>and I2R losses? If the size matters, where would the point of indecision be? | | What I'd do is run 120/208V to the data center, however I'd try to | configure as much as possible to run off 208V, wired line to line, with | everything balanced equally between each of the 3 possible pairs of phases. | Lots of 6-15P cords and 6-15R or 6-20R recepticles. | Why? Most computer power supplies will autoswitch between anything from | 100-250V, but the real reason is they generate harmonics which tends to | overload the neutral when large numbers of them run off 3 phase L-N (120V). | If they're wired line-line, there's no neutral current, plus the power | spike (which when wired L-N happens at the peak voltage) becomes two | smaller power spikes on either side of the waveform peak (the peak between | the phases).

Connecting L-L does not eliminate the harmonic issue.

You will see up to a 15% current increase in the worst case (that of the load being entirely triplens) on the hot phases (because the normal case of phases balancing on shared hots reduces the sinusoidal current by 15%).

But the big surprise will come at the neutral interconnection point in the system source transformer, assuming it has a WYE secondary (e.g. 120/208). And if the primary is delta, it will still have circulating currents.

You still have to address the harmonic issue. Connecting L-L means you only have to derate a branch circuit by up to 15% in the worst case. The rest of the issues are still present.

| There will be plenty of 120V power available, of course, for devices which | do not have 120/240V autosense, things with wall warts, and whatever widget | someone may need in the future that runs off "normal" power. | Any of these with a switching PS will generate neutral harmonics, but | (hopefully) not enough to overload, plus their line peak will happen at | a different point of the line waveform than the big load.

Modern PSUs these days are supposed to have a lot less harmonic current. So in the case of L-L you probably won't see anywhere near the 15% worst case scenario, and won't have to derate the branch circuits at all. But if the entire transformer load is all PSUs, you might want to look at just how much oversizing or K-factor your transformer has for handling harmonics.

I never said it did. I said it eliminated the neutral harmonic load (no neutral current, at least until you start using 120V devices) and spread out the line pulses.

A 15% increase worst case is much better than 3 times load worse case! (neutral harmonic current)

Yes the supplying transformer will have to be properly rated for harmonics. No change there.

Where I used to work there was a big box much like a transformer which was a harmonic neutralizer of some sort. One of these might be necessary for larger data centers.

Again: 15% or 3 times?

On Wed, 3 Dec 2008 17:43:56 +0000 (UTC) Michael Moroney wrote: | snipped-for-privacy@ipal.net writes: | |>On Wed, 12 Nov 2008 23:54:16 +0000 (UTC) Michael Moroney wrote: |>| snipped-for-privacy@ipal.net writes: |>| |>|>If you are building a data center for a large number of computers (lots of PCs |>|>and server sized PCs, with all the network infrastructure), and this involved |>|>a large amount of power, and this were in a location where the standard user |>|>line-to-neutral voltage is 120V, but you could get a line-to-neutral voltage |>|>of 240V by special order, would you choose the latter to reduce wiring costs |>|>and I2R losses? If the size matters, where would the point of indecision be? |>| |>| What I'd do is run 120/208V to the data center, however I'd try to |>| configure as much as possible to run off 208V, wired line to line, with |>| everything balanced equally between each of the 3 possible pairs of phases. |>| Lots of 6-15P cords and 6-15R or 6-20R recepticles. |>| Why? Most computer power supplies will autoswitch between anything from |>| 100-250V, but the real reason is they generate harmonics which tends to |>| overload the neutral when large numbers of them run off 3 phase L-N (120V). |>| If they're wired line-line, there's no neutral current, plus the power |>| spike (which when wired L-N happens at the peak voltage) becomes two |>| smaller power spikes on either side of the waveform peak (the peak between |>| the phases). | |>Connecting L-L does not eliminate the harmonic issue. | | I never said it did. I said it eliminated the neutral harmonic load | (no neutral current, at least until you start using 120V devices) | and spread out the line pulses. | |>You will see up to a 15% current increase in the worst case (that of the |>load being entirely triplens) on the hot phases (because the normal case |>of phases balancing on shared hots reduces the sinusoidal current by 15%). | | A 15% increase worst case is much better than 3 times load worse case! | (neutral harmonic current)

Yes, way better. But not zero.

|>But the big surprise will come at the neutral interconnection point in the |>system source transformer, assuming it has a WYE secondary (e.g. 120/208). |>And if the primary is delta, it will still have circulating currents. | | Yes the supplying transformer will have to be properly rated for harmonics. | No change there. | | Where I used to work there was a big box much like a transformer which was | a harmonic neutralizer of some sort. One of these might be necessary | for larger data centers.

Probably a zig-zag transformer.

|>You still have to address the harmonic issue. Connecting L-L means you |>only have to derate a branch circuit by up to 15% in the worst case. The |>rest of the issues are still present. | | Again: 15% or 3 times?

Just don't assume 0%.