Author Topic: Designing ATX Power Supply when the rated DC voltages are already available  (Read 11737 times)

0 Members and 1 Guest are viewing this topic.

Offline hkBattousaiTopic starter

  • Regular Contributor
  • *
  • Posts: 117
  • Country: 00
I have an efficient, high-power and reliable power supply. I want to use this power supply to feed as many PCs as possible. My aim is to reduce fan noise, stop cleaning dust clogged heat-sinks before summers and increase power efficiency. I want to design interfacing circuits for each PC as seen in the schematic at the bottom.

Can you please answer the questions below.
  • Are the MOSFETs used correctly?
  • What kind of input filter should I use? (The power lines may travel large distances (~50m) to reach furthest computers.)
  • Do I need output filters?
  • Will this circuit work if I correct any faults you will mention?
  • What improvements can I add?

 

Offline Mike Warren

  • Supporter
  • ****
  • Posts: 437
  • Country: au
    • Personal Website
50M at the sorts of currents PCs use is going to need massive gauge wiring to have a low enough voltage drop not to cause problems. I can't see this being practical at all.
 

Offline mij59

  • Frequent Contributor
  • **
  • Posts: 693
  • Country: nl
To keep the voltage drop low use something like this
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 16385
  • Country: za
You could use point of use regulation, then you only need a 12V supply to feed to each system, doing the regulation to 5V and 3V3 at the motherboard connector. That can come from a large PSU directly, as most high power ones make a single ( or multiple) 12V supply internally then use converters for the other voltages.

http://www.ebay.com/itm/12V-160W-Pico-ATX-switch-PSU-Car-Auto-MINI-ITX-ATX-Power-Supply-24pin-DC-ATX-/261654849070?pt=LH_DefaultDomain_0&hash=item3cebd8262e

Something like this if the power rating is right.
 

Offline jwm_

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: us
    • Not A Number
Modern servers are designed with this in mind, they generally draw almost  all their power from the 12v line so the other lines can be generated locally by small switchers that don't need to be very big. A usual setup is to have one beefy 12v supply feeding bus bars for a whole rack of servers. Since you are making a board anyway, including local generation of -12 and -5 will be easy and save you copper, and if your servers follow the 12v policy you can do 5 and 3.3 locally too.

The old server power supplies can be gotten cheaply (15 bucks)on eBay and are quite high quality high amp 12v supplies (I have repeatedly shorted one with no damage, precautions were taken against arc flash of course) and can generally be converted to be floating.

Offline Whales

  • Super Contributor
  • ***
  • Posts: 2122
  • Country: au
    • Halestrom
Optimisation note: -5V rail was dropped from the ATX spec years ago and very, very few things use -12V either.  Unless you are running very old computers you will need neither.

As SeanB suggests local regulation from a single feed supply would be the neatest solution, especially if you send over 12V (eg 20V) and locally regulate.  This way you don't have to worry about the vdroop over the wiring as much (50M is a long distance for these voltages and current levels).  You may find 50M an impossible feat with your current design, depending on how much you are willing to spend on wires (it'll probably be much more than the energy savings for quite a while).

Offline hkBattousaiTopic starter

  • Regular Contributor
  • *
  • Posts: 117
  • Country: 00
Optimisation note: -5V rail was dropped from the ATX spec years ago and very, very few things use -12V either.  Unless you are running very old computers you will need neither.

How much current does a typical motherboard draw from -5V and -12V rails? If the negative current requirement is very low, maybe I can use simple charge pumps to generate negative voltages, can I? What could be the worst case current requirement?
 

Offline macboy

  • Super Contributor
  • ***
  • Posts: 2309
  • Country: ca
50M at the sorts of currents PCs use is going to need massive gauge wiring to have a low enough voltage drop not to cause problems. I can't see this being practical at all.
I agree. 30 A on the 12 V line is not unusual, and a gaming rig will double that. At 30 A, you will need at least 1/0 gauge cable to keep the voltage drop below 2 V (10 V at the end) over 100 m of cable (50 m each hot and ground). Better would be 2/0 gauge for 1.5 V total drop, and even that is pushing it. The conductor in that will be 10 mm across, and 100 m of it will weigh over 100 lb (over 50 kg). You can buy this as welding cable, and it will cost you ~$3 per foot, or ~$1000 for your 100 m (50 m x2) run. Have fun with that.
 

Offline RR

  • Contributor
  • Posts: 30
  • Country: cz
My aim is to reduce fan noise, stop cleaning dust clogged heat-sinks before summers and increase power efficiency.
PSU usually are not the main source of noise in system and you still dont get away from cleaning dust.
Eff will also suck with voltage drop over the long wires.

How much current does a typical motherboard draw from -5V and -12V rails?
-5V is long gone. -12V guess if its even used then max to power couple of opamps so guess <500mA

usefull link - http://www.formfactors.org/developer/specs/atx12v_psdg_2_2_public_br2.pdf
 

Offline jwm_

  • Frequent Contributor
  • **
  • Posts: 319
  • Country: us
    • Not A Number
Hopefully you are not using pentium-4 era machines, those can draw 100A on the 3.3V line. It was getting pretty absurd, hence the switch to 12V only.

I don't think there is any way you can distribute 5 and 3.3 effectively over any distance and there isn't much point. If a lot of power, you are going to get a signifigant voltage loss, 1V drop on 12V hurts a lot less than 1V on 3.3V and if it is low power than a small, inexpensive buck converter can be used at point of use. you certanily can do -12V locally, the tolerances required by ATX are really wide for it. unregulated charge pump will probably do.

I think the best way (if you really want DC) is to use one of those little power supplies meant to be used for a car posted earlier at each node and shorter segments. And if you find your wires are too long, just lug a 12V marine deep cycle battery back and forth to your charging station occasionally and run them off that.

The ATX standard also requires two independent 12V rails with independent current limiting, you probably can get away without that, but good to be aware of.

Note that some cases/motherboards expect the ATX supply airflow to occur in a specific spot. you will want to make sure it isn't overheating without the supply as you are giving up that extra circulation too.

Offline richcj10

  • Supporter
  • ****
  • Posts: 201
  • Country: us
Re: Designing ATX Power Supply when the rated DC voltages are already available
« Reply #10 on: February 17, 2015, 08:21:23 pm »
50M at the sorts of currents PCs use is going to need massive gauge wiring to have a low enough voltage drop not to cause problems. I can't see this being practical at all.
I agree. 30 A on the 12 V line is not unusual, and a gaming rig will double that. At 30 A, you will need at least 1/0 gauge cable to keep the voltage drop below 2 V (10 V at the end) over 100 m of cable (50 m each hot and ground). Better would be 2/0 gauge for 1.5 V total drop, and even that is pushing it. The conductor in that will be 10 mm across, and 100 m of it will weigh over 100 lb (over 50 kg). You can buy this as welding cable, and it will cost you ~$3 per foot, or ~$1000 for your 100 m (50 m x2) run. Have fun with that.

errrrrr wrong.....

@ 8 Gauge the Ohms/foot is .0006
That means total resistance of 0.062 @ 30A = 1.88 < 2v drop. You don't need 2/0, 1/0, etc to do this task.
 

Offline suicidaleggroll

  • Super Contributor
  • ***
  • Posts: 1453
  • Country: us
Re: Designing ATX Power Supply when the rated DC voltages are already available
« Reply #11 on: February 17, 2015, 08:34:35 pm »
errrrrr wrong.....

@ 8 Gauge the Ohms/foot is .0006
That means total resistance of 0.062 @ 30A = 1.88 < 2v drop. You don't need 2/0, 1/0, etc to do this task.

That's only one side, you need to get the current back to the supply, so you're up to 3.8v drop.  Actually nevermind, I'm not sure what you're doing.  50m is 328 feet round trip, that's 0.2 ohms, and a 6.1v drop at 30A.

That's also only one computer using a measly ~360W.

The whole idea is pretty pointless.  As was said above, the power supply is usually the quietest part of a machine, and since it blows the air OUT of the case, not IN, is not the one primarily responsible for dust getting stuck in your heatsinks.  You should be looking at the intake fans, usually located in front of the hard drive tray to keep them cool.  Maybe put a filter on it?  If you want the computer to be quieter, then use bigger fans running at a slower speed.  Replace your CPU and chassis fans with 120mm with some little rubber dampeners for attaching them to the chassis and the machine will be nearly silent.
« Last Edit: February 17, 2015, 10:26:41 pm by suicidaleggroll »
 

Online ajb

  • Super Contributor
  • ***
  • Posts: 2786
  • Country: us
Re: Designing ATX Power Supply when the rated DC voltages are already available
« Reply #12 on: February 17, 2015, 08:38:38 pm »
Do telco racks still generally run on 48VDC, or have cheap inverter-based UPSes made that obsolete?  Presumably there would be ATX power supplies designed for 48V input designed for that market. 

Anyway, if you distribute something like 48VDC, you have plenty of headroom for regulating down to 12V even after accounting for voltage drop, which will be only 1/4th of what it would be at 12V (and your I2R losses drop to 1/16th).  At 48V, you also still have the option of easily doing a directly-connected battery bank for backup, although if you're going that route 24V might be a better compromise in terms of being able to use lower cost off-the-shelf battery management hardware.
 

Offline Whales

  • Super Contributor
  • ***
  • Posts: 2122
  • Country: au
    • Halestrom
Re: Designing ATX Power Supply when the rated DC voltages are already available
« Reply #13 on: February 17, 2015, 10:18:35 pm »
@ 8 Gauge the Ohms/foot is .0006
That means total resistance of 0.062 @ 30A = 1.88 < 2v drop. You don't need 2/0, 1/0, etc to do this task.

Even 1.88V on the 12V rail is way too large of a drop to stay within ATX spec.  This is more of a problem than you think for most computers and can lead to instability or (more commonly) motherboard chipsets refusing to power-on because power is out of spec.

Quote
The ATX standard also requires two independent 12V rails with independent current limiting, you probably can get away without that, but good to be aware of.

Muahaha *scoff* safety features.  If you can't arc weld without over-current protection tripping, then what's the point? 

Offline Phaedrus

  • Frequent Contributor
  • **
  • Posts: 714
  • Country: us
Re: Designing ATX Power Supply when the rated DC voltages are already available
« Reply #14 on: February 17, 2015, 10:24:23 pm »
Just a note, the requirement for multiple rails was actually part of the EPS12V server PSU standard, and was quickly made optional as it was poorly written and caused a lot of companies a lot of headache, as the standard required the CPU to have its own dedicated +12V rail, and all other +12V hardware (including, importantly, the graphics card) to share the second rail. With the advent of high-end graphics cards with >200W power consumption, this caused many gaming systems to crash when used with "multi rail" power supplies, as they tripped the over current protection and the PSU shut down. This led company PC Power & Cooling to start marketing "single rail" power supplies with their +12V OCP circuit removed entirely. Later manufacturers copied this trend, and to this day many users are gun shy about multi rail PSUs, despite them being significantly safer than single rail units.
"More quotes have been misattributed to Albert Einstein than to any other famous person."
- Albert Einstein
 

Offline Whales

  • Super Contributor
  • ***
  • Posts: 2122
  • Country: au
    • Halestrom
Re: Designing ATX Power Supply when the rated DC voltages are already available
« Reply #15 on: February 17, 2015, 11:14:13 pm »
A lot of the time PSU makers provide enough juice on each split rail for it to not matter anyway, but opinions are indeed still divided. 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf