Maker Pro
Maker Pro

Wire ATX Power Supply Busbar for PC Farm (12V 256A total)

alexhk

Jun 17, 2012
3
Joined
Jun 17, 2012
Messages
3
Wire ATX Power Supply Busbar for Atom CPU Farm (12V 256A total)

Hello

I am new here and as you can probably tell very soon, I don't have a lot of experience in electronics. I am a programmer with some basic electronics understanding. To make matters worse, English is not my native language so please be gentle ;) (I am in Asia but I was born in Europe)

For a small project I have been using PC Power Supplies (ATX PSUs) to power 16 small low-cost Atom mainboards (they are basically small noname mainboards with an Intel Atom CPU and everything on board, such as display card, wifi, etc). They use 12V power inputs and have a low power consumption (around 40W continues).

Scroll down for a photo of these Atom boards (I am not using these, but you get the picture).

The ATX PSU I have been using has a single 12V rail that can provide up to 816W/68A with a very good efficiency (Cooler Master M2 Silent Pro 850W is the type). I am actually only using around 640W (16 Atoms * 40W), but to have a safety buffer and for the easy of calculation, I would like to assume we are talking about 4A per Atom Board (48W).

What I did so far is wiring 16 wires of 16AWG from the +12V output pins of the power supply to the 16 different Atom Boards. Same for the GND pins. I've been using this setup for a few weeks now, it works fine and looks professional.

But now I have to increase the no. of Atom PCs to 64 in total, so I will need 4 PSU's in total. To make matters worse, I have some space constraints so I can't wire one by one. I will have to go to a busbar outside the case, then have a +12V and GND wire in 8AWG go into a case, and then go to another busbar for power distribution again.

And this part is exactly where I hope to get help here. We are talking about a lot of Ampere and really fat cables for my taste, so I just want to double check what I am doing won't burn the house down and I don't kill myself in the process.

So what I want to do is (once for +12V and once for GND):

Code:
[ATX PSU]

  IIIIIIII                (+12V 8 wires of 16AWG? enough?)
   (1 foot)
  IIIIIIII
  
[BUS BAR]

     I                   (+12V 1 wire 8AWG in conduit)
   (3 foot long)
     I

[BUS BAR]

  IIIIIIIIIIIIIIII           (+12V 16 wires 16AWG)
  (1 foot)
  IIIIIIIIIIIIIIII          

[16x Atom]
Repeat for GND.

So after I am done I will have one ATX PSU that powers 16 Atoms again.
On the PSU side I have 2 busbars, one for 12V one for GND.
The two 8AWG wires go through a cable conduit.
On the farm side I will have 2 more busbars.
Then I have the 16 pairs of 12V and GND again to the Atom PCs.

Here are my questions:

I will have 4 of these setups, independent from each other. But the fat ~8AWG (~10mm²) wires (4x 12V and 4x GND) will run in the same conduit into the case. Is this a problem?

While I initially used to connect 16 wires to the PSU, my busbars don't have 16 holes. I am using 16AWG wires and the sockets on the PSU are rated high enough. How many 16AWG wires in parallel would you run from the PSU to the busbar?

I am planning to use bootlace ferrules and heat-shrink tubing on the 8AWG wires and screw them down to the bus bars real good. Is this "good enough" even for the 64A current on them?

Anything else I should consider when I am doing all this? (except getting life insurance, being extra careful, and having somebody double check my work)

Here are the pictures:

atoms-0001L.jpg


Here is what I wired up so far (busbar rated 128A):
photo%25202.JPG
 
Last edited:

Harald Kapp

Moderator
Moderator
Nov 17, 2011
13,700
Joined
Nov 17, 2011
Messages
13,700
AWG8 is nominally good for up to 73 A when used in chassis wiring. And running four of the wires in parallel neither is a problem.
However, at 3 feet each (+12V, GND) the wire resistance of the short piece within the conduit will be 0.0037692 Ohm. This doesn't sound much, but at 64 A (4A*16 PCs) this will give a voltage drop of 0.2412 V and a power loss of respectable 15 W! Add the voltage loss at the non-ideal contact points from the busbar to tha cable and vice versa and you'll have lots of losses. I therefore recommend using more or thicker wire. If you double the number of wires (of the wire diameter) you halve the losses.

See here: http://www.powerstream.com/Wire_Size.htm

Harald
 

alexhk

Jun 17, 2012
3
Joined
Jun 17, 2012
Messages
3
Thanks a lot Harald, that is very helpful info. Coincidentally we also speak the same native language, I was born in the alpine nation (Austria) to your south ;)

And running four of the wires in parallel neither is a problem.

I will be running 8 wires in the same conduit, 4x 12V and 4X GND. But it won't be 'in parallel' electrically speaking though. It will be 4 different circuits of the 4 power supplies.

At 3 feet each (+12V, GND) the wire resistance of the short piece within the conduit will be 0.0037692 Ohm. This doesn't sound much, but at 64 A (4A*16 PCs) this will give a voltage drop of 0.2412 V and a power loss of respectable 15 W!

I am not at the office right now, but the PSU actually outputs a bit more than 12V (12.10V if I am not wrong). After putting together the bus bar in the picture, I have measured and the voltage drop fortunately wasn't as big as the calculation would suggest (I'll update with real numbers once in the office).

My Atom PC's can accept a range of voltage inputs so that shouldn't be a problem. 12V is the most efficient though.

Once the whole assembly is up and running and the Atom's connected, I can actually measure the power consumption on the wall plug. I still have the numbers for the direct connection (16 Atom's wired directly), so comparing the two and knowing the loss will be easy.

I therefore recommend using more or thicker wire. If you double the number of wires (of the wire diameter) you halve the losses.

Is this recommendation for the "big" wires between the bus bars (3 feet) or do you recommend to run more wires into the bus bar from the PSU itself?

The cable I called 8AWG (8.36mm²) above will probably be a cable with a cross sectional area of 10mm², since this is the one I have in stock and I couldn't find 8AWG here at my location easily. That's already 1.64mm² thicker than in my initial plan (I'll still try to get a black one though, I don't fee comfortable wiring 12V and GND in the same yellow color).
 
Last edited:

CocaCola

Apr 7, 2012
3,635
Joined
Apr 7, 2012
Messages
3,635
I don't fee good wiring 12V and GND in the same yellow color.

Just spiral wrap some electrical tape around it...

OR

Keeping the cut and exposed ends above water (so you don't get water inside the wire) drop it in a bath of boiling RIT or equivalent fabric dye, a lot of wire insulations will take the dye, it might not be a perfect dye job but it should shift color... Use blue and you might even get a nice shade of green for ground... Of course not all insulation will take the dye so you might want to test first, I have had some good luck in the past...
 

alexhk

Jun 17, 2012
3
Joined
Jun 17, 2012
Messages
3
AWG8 is nominally good for up to 73 A when used in chassis wiring. And running four of the wires in parallel neither is a problem.
However, at 3 feet each (+12V, GND) the wire resistance of the short piece within the conduit will be 0.0037692 Ohm. This doesn't sound much, but at 64 A (4A*16 PCs) this will give a voltage drop of 0.2412 V and a power loss of respectable 15 W! Add the voltage loss at the non-ideal contact points from the busbar to tha cable and vice versa and you'll have lots of losses. I therefore recommend using more or thicker wire. If you double the number of wires (of the wire diameter) you halve the losses.

See here: http://www.powerstream.com/Wire_Size.htm

Harald

Hi Harald,

I just wanted to give you an update.

I have been running this setup for a while now, with 8 wires each 10mm² (AWG8 not available where I am, in Asia). It's 4 different circuits, 4x 12V and 4x GND, with 64A each.

It worked, but the wires are getting hot where I feed them through a short pipe. I put a thermal probe between the cables and it's more than 60°C (140° F). Attached a picture of this setup.

I also checked the cable manufacturer data sheet. The manufacturer says the cables are good for up to 72A, but when laying multiple-lines I need to multiple by a factor of 0.49 for 7~15 cables. This means the cable is actually only good for 35.28A? (72*0.49)

I am now moving up to 16mm². The cable says good for 98A, corrected by 0.49 results in 48A. That is really disappointing, but I am still going to try it, it should not get as warm as the old one.
 

Attachments

  • wire-pipe.JPG
    wire-pipe.JPG
    32.1 KB · Views: 293
Top