Power Factor Correction and Servers

[Halcyon]

I noticed an option in the BIOS for a newer Dell server, that allow you to turn PFC on or off. It's not something you usually see in this type of equipment.

Why would someone want to turn PFC off? Does it have some kind of drawback that I'm not aware of?

[Monkeh]

Apparently it increases standby power consumption. I'm not quite sure who has a use-case for large numbers of servers sat powered off, but apparently someone wanted it enough for Dell to do it..

[NiHaoMike]

Having PFC on increases power use at light load. Some UPS systems perform the PFC function and since the UPS would be rated for the maximum VA draw, the increased current at low load would not be a problem.

That said, it probably would be more accurate to make the choices "Auto" and "Always On" since PFC will always have to operate under significant load.

[Miyuki]

Either a light load (what should be solved automatically by the PFC controller)
Or a UPS issue. This sounds more plausible.
You can even decide to use non-sinusoidal output on UPS as this will increase efficiency. But most PCF controllers will go crazy with it.
There are even some UPS systems that provide DC, but it tends to be on lower voltages as it is a hustle to fuse or disconnect 300-400V DC
It is a pure waste to make UPS for servers to produce pure sine as nothing there use AC and they you have to convert it against back to DC with PFC
//edit
But the manual says it decrease consumption in Sleep mode

[Marco]

I see minus 48v as server power, why minus?

[woodchips]

In the very old days, a telephone exchange ran from -50V supply.

[Miyuki]

I looked at DC power supplies and most rack mount servers have a DC option of nominal 48V (40 V to 75 V), sometimes marked negative. Wonder if they are isolated or if they use one pole as chassis ground
This is also a default POE voltage level.

[Circlotron]

In the very old days, a telephone exchange ran from -50V supply.

This is to prevent all your miles and miles of telephone wires corroding. For some reason it is the positive side that corrodes. Have a look at your car battery terminals. This is also why older British made cars had positive connected to the body work. The body has far greater surface area than the car wiring so the corrosion on it would be minimal compared to if the the wiring were positive like it is nowadays.

[langwadt]

In the very old days, a telephone exchange ran from -50V supply.

This is to prevent all your miles and miles of telephone wires corroding. For some reason it is the positive side that corrodes. Have a look at your car battery terminals. This is also why older British made cars had positive connected to the body work. The body has far greater surface area than the car wiring so the corrosion on it would be minimal compared to if the the wiring were positive like it is nowadays.

positive chassis would be convenient for low side switching with a mosfet

[Red Squirrel]

I see minus 48v as server power, why minus?

Has to do with corrosion as others mentioned.  Telephone exchanges run on strings of 24 lead acid batteries which is 48v nominal, 54v float.  The positive is grounded.  The telephone equipment runs straight off the -48v, including DSLAMS etc.    The rectifiers (AC DC power supplies essentially) keep them charged up and if power goes out the rectifiers shut off but everything else just keeps going off the batteries.  When voltage starts to get around 47v then they'll hookup a generator to top up the batteries.  I've seen equipment run on as low as 44 volts but we avoid letting batteries get that low as it's bad for them.

As for PFC I heard it can cause issues with UPSes as most off the shelf UPSes are square wave and PFC does not like this, so I guess if it's causing an issue you would be able to turn it off.  First time I hear of that being an option myself though.

Been wanting to upgrade my UPS system to dual conversion 48v similar to a telco setup, but it's one of those things I don't really want to spend the money on either.  Looking at around 5 grand or so by the time I buy the PSUs (can't find anywhere that sells rectifiers), inverters and batteries.   Was looking at some Meanwell PSUs, some support paralleling so I'd probably go that route, then one inverter per PDU.   Maybe at some point I will build my own power electronics but not sure if I trust myself enough to run all that equipment off a DIY job.

[Marco]

Had to do with corrosion at this point, POTS is long dead and servers are an entirely different market. The moment -48V leaves the insulated copper wires it's DC-DC'd to positive voltage rails, it can't provide impressed current cathodic protection for anything relevant for servers.

[Bobson]

POTS is long dead and servers are an entirely different market.

Not always. NEBS compliant servers/routers/switches have POTS legacy. Also, telcos consume very high share of servers.

[NiHaoMike]

In the interest of prolonging the life of existing lines, wouldn't it make sense to have a protocol where the customer end equipment tells the ISP that it's self powered and no DC bias is required, keeping it that way as long as there's still an active link?

There also exists a standard for supplying power from the customer end to the ISP, in order to improve resilience to natural disasters.
https://www.microchip.com/en-us/products/power-management/reverse-power-feed-rpf

[Red Squirrel]

POTS is still highly used, but I'd say we're kind of in the transition phase.  There is no reason to roll out new POTS but also no reason to pull out existing that still works.  What ends up happening is even the newer CO equipment that powers fibre optic, cellular etc is designed to run on -48v because that's what the power plant in the CO is setup as.

I think there is also a DC standard for data centres that is in the 100's of volts.  I could see telecom eventually switch to that once POTS is out since 48v is expensive in copper.  000000 cable is not uncommon within the building.  The price of copper now days makes new runs of that stuff pretty much unfeasible.  The cables were originally installed at a time when copper was cheap.

[asis]

Hi,
This is most likely due to the problem of some UPSs going into current protection at the moment the power supply of the server or several servers is turned on at once, since transients in PS with PFC create large currents at the moment of turning on.
Moreover, if a switch to a diesel generator is made, PFC can play a cruel joke.
The change in the sinusoidal shape of the input voltage forced all  manufacturers to use switching power supply with PFC.
Here are some highlights of how PFC works.