Overclock.net banner
1 - 10 of 10 Posts

·
Registered
Joined
·
1,190 Posts
Discussion Starter · #1 ·
WARNING: I do not encourage anybody to do this at home! If you imitate this PSU mod, you do this at your own risk!

Results:

After a bit of experimenting, I assessed that this would be the recommended minimum to power a Thin mITX board with an ATX12V power supply:

Cables from PSU:
  • PS_ON# (green)
  • PWR_OK (gray)
  • 5VSB (purple)
  • GND (black)
  • 12V (yellow)
  • GND (black)
Additionally, when using a GPU with auxiliary power connectors:
  • 12V * 2+ (yellow)
  • GND * 2+ (black)
Mods to the PSU:
  • Permanently connect one 3.3V wire (orange) to 3.3V sense (brown)
  • Solder a resistor between one 5V wire (red) and one ground wire (black), 220Ohm worked for me.
  • Shorten all cables not needed so you can isolate them and tuck them inside the PSU

More information will follow! For more details, see the OP and the other posts in this thread!

OP:

So, this may be a bit of an odd one, but I've got this idea stuck in my head for two months now. It's a long read, but it was a really interesting thing to do.

I wanted to find out if you can power a Thin mITX board with a PSU that adheres to the ATX standard, which would allow to reduce cable clutter by not having a 24pin ATX connector, maybe not even a SATA power cable, and space for an HDD directly above the mainboard (there's ~73mm of space between the I/O shield and an NH-L9i on every Thin mITX board). Basically, it would make very small builds with integrated PSUs possible.

Theory:

Because the Gigabyte Thin mITX boards with a 1150 socket have a wide input range of 12V-19V instead of the regular 19V (which I actually tested before), this becomes a feasible plan, we can use the 12V a regular PSU puts out to power one of those mainboards while reducing the number of cables needed for the power connector from 23 to 2. Noice.

But: ATX PSUs have two signal lanes, called PS_ON# and PWR_OK. The first one is used by the mainboard to turn the PSU on, the second one indicates whether the 3.3V and 5V lanes are outputting power as the specification requires.
Thin mITX boards have voltage regulators for 5V and sometimes 3.3V on board, so let's ignore the latter issue for now.
But the PS_ON# signal is a bit tricky: As long as that signal is held to ground, the PSU is turned on and will stay turned on. But as soon as the connection from this lane to ground is cut, the PSU will be turned of immediately, so we would have to use a maintained switch instead of a regular one and we would cut power to the board every time we open the switch to turn the PC of.
We can get around that by using a microcontroller to emulate the behaviour of an ATX12V mainboard, in my case an Arduino.
Turning the PC off would then require the Arduino to first send a signal to the mainboard to turn it off, wait for that and then turn off the PSU.

Fortunately, ATX12V PSUs also have a 5V Standby supply which is always on when the PSU is connected to AC power. The Arduino does only have a 7V-12V input, but it also has a USB port, wich operates on 5V, so we can actually have the arduino on all the time and do cool stuff like a breathing power LED when the PC is turned of, for example.

Practical Experimenting:

WARNING: I do not encourage anybody to do this at home! If you imitate this PSU mod, you do this at your own risk!

I already got my hands on a GA-Q87TN and an Arduino Uno R3 and today, the FSP270-60LE finally arrived at my doorstep. It is a very interesting PSU in that it not only is a short FlexATX PSU, but also that one of the two 12V rails is rated as 10A, while the other one is rated as 16A. So all components on the mainboard can only consume 120W, but a GPU would be allowed to take as much as 192W, which is enough for an overclocked 980. For gaming, this is pretty interesting.



Also note the non-standard mounting flange and the very short ATX 24pin cable.

Stage 1: Powering the Arduino
I stripped a USB cable and connected its 5V wire (red) to 5VSB (purple) of the ATX 24pin and ground to ground. Plugging in the Arduino to the USB cable and the PSU into AC already supplies the Arduino with power and we can run programs on it, provided we loaded them onto the Arduino from another PC first.



Stage 2: Turning on the PSU
Then I connected the PS_ON# lane to one pin of the Arduino which I set as OUTPUT. Added a button and an LED on the Arduino and a multimeter at the 12V output of the ATX 24pin for debugging, wrote some code and now I can turn on the PSU by pushing a button on the PSU. The fan starts spinning and the multimeter shows 11.78V on the 12V lane. A bit low for my taste, but inside the 5% tolerance specified by the manufacturer.



Stage 3: Problems
So now I thought everything would be fine and dandy, I disconnected the Multimeter, connected the mainboard in its place, turned the PSU on and shortened the power button pins on the board. The fan of the NH-L9i spun for a brief moment, then it and the PSU turned off. The Arduino was still running, but the 5VSB lane is unaffected by pretty much everything so that didn't help a lot.
I decided to connect a second, green LED between the PWR_OK signal and ground from the PSU. (Don't forget the resistor)
The results were exactly as expected, the green LED turned on a little after the red one and they turned off at the exact same time.

But for some reason, the PSU kept turning off even without me doing anything, and the red LED did so, too. Something isn't right, I'll look at it tomorrow.

That's it for now, my brain can't think anymore. I'll try to solve these issues tomorrow, but I'm already quite happy with how things are going. Hopefully I'll be able to power a GPU with this as well, that would open up so many possibilities!
 

·
Registered
Joined
·
137 Posts
Really interested to see how you work this out. I've been pondering the same issue for a while as I'd like to hook up my new i7 NuC to a GTX 970 via an M.2 connector. I was thinking about using a SWEX from BPlus to handle the power on issue (pretty sure the 2 pin jumper is for a power switch) and then running lines to the 4 pin auxiliary power header on the NuC board. NuC uses this pinout:

Pin Descriptive Name
1   5V
2 Power draw sensor
3 Ground
4 12V - 19V DC input

Would this solution work for the 2pin Minifit connector on the Thin ITX board? Given the right ATX supply, this would leave the 6 pin/molex/etc. lines for powering your GPU and hard drive as well as any other bits and pieces.

This is where I found info on the SWEX:

http://forum.techinferno.com/enclosures-adapters/8441-%5Bguide%5D-switching-atx-psu-using-paperclip-trick-swex.html
 

·
Registered
Joined
·
103 Posts
Arduino's can do strange things sometimes. perhaps on startup the 5VSB spikes or dips causing the Arduino to crash?
What you can try is use a simple MOSFET and drive it with the arduino, thus causing a short between the PS_ON and GND of the PSU.

This is how I power on the HDPLEX in my build. I use the 12V, or 5V in your case. to switch a MOSFET on that shorts the PS_ON to GND. Simple and very compact.

Like this:
medium.jpg

Only difference is the +12V would be the 5V from you Arduino.

Also add a resistor between Gate andGround to pull down the line when the arduino is off or starting doing funny stuff.
 

·
Registered
Joined
·
1,190 Posts
Discussion Starter · #5 ·
Quote:
Originally Posted by Aibohphobia View Post

Really neat project, I hope you get it figured out because this would open up some interesting build possibilities.
Thanks! Yeah I already modelled a ~4L case with a GTX 970, this PSU and a thin mITX board inside, it even fits a 2.5" HDD
biggrin.gif

Quote:
Originally Posted by ilovelampshade View Post

Really interested to see how you work this out. I've been pondering the same issue for a while as I'd like to hook up my new i7 NuC to a GTX 970 via an M.2 connector. I was thinking about using a SWEX from BPlus
to handle the power on issue (pretty sure the 2 pin jumper is for a power switch) and then running lines to the 4 pin auxiliary power header on the NuC board. NuC uses this pinout:

Pin Descriptive Name
1   5V
2 Power draw sensor
3 Ground
4 12V - 19V DC input

Would this solution work for the 2pin Minifit connector on the Thin ITX board? Given the right ATX supply, this would leave the 6 pin/molex/etc. lines for powering your GPU and hard drive as well as any other bits and pieces.

This is where I found info on the SWEX:

http://forum.techinferno.com/enclosures-adapters/8441-%5Bguide%5D-switching-atx-psu-using-paperclip-trick-swex.html
I guess it would work, but why bother with this huge 24pin connector when I could just use an Arduino nano instead and eliminate all the clutter of cables I don't need plus gain a few features like an animated power LED? I've already gone down from 23 cables to 8 now and 6 will probably suffice. Everything else is a waste of space IMO.
 

·
Registered
Joined
·
137 Posts
I obviously wasn't clear on the intent of this mod. So what you're essentially trying to do is route power through the arduino to handle on.off plus some other potential features while trimming all unnecessary cables from the PSU so you only need to power the Mobo and GPU?
 

·
Registered
Joined
·
1,190 Posts
Discussion Starter · #7 ·
Quote:
Originally Posted by QinX View Post

Arduino's can do strange things sometimes. perhaps on startup the 5VSB spikes or dips causing the Arduino to crash?
What you can try is use a simple MOSFET and drive it with the arduino, thus causing a short between the PS_ON and GND of the PSU.

This is how I power on the HDPLEX in my build. I use the 12V, or 5V in your case. to switch a MOSFET on that shorts the PS_ON to GND. Simple and very compact.

Like this:
medium.jpg

Only difference is the +12V would be the 5V from you Arduino.

Also add a resistor between Gate andGround to pull down the line when the arduino is off or starting doing funny stuff.
Writing LOW to an output pin does pretty much the same thing (pulling a signal to ground), but pull-down resistor is a good buzzword. There were different problems though, as this update will show you.
wink.gif

Quote:
Originally Posted by ilovelampshade View Post

I obviously wasn't clear on the intent of this mod. So what you're essentially trying to do is route power through the arduino to handle on.off plus some other potential features while trimming all unnecessary cables from the PSU so you only need to power the Mobo and GPU?
I don't route power through the Arduino, I use it to turn the PSU on and off, but the other stuff is correct, yes.

PROGRESS!

Stage 3.5: Fixing things, booting

So for some inexplicable reason, the PSU turned on and off seemingly randomly. The problem was that I forgot to use a pull-down resistor on the main button.
doh.gif
.
After adding that, everything worked as expected and the PC was entering BIOS just fine. Phew.



Or so I though. Curiously, when I started the PC, the fan of the PSU stopped spinning which I previously perceived as the PSU turning off. This wasn't due to any semi-fanless operation mode as it was about to turn out.
So in order to boot, I got the mSATA SSD from my main rig and put it into the GA-Q87TN, reached the loading screen of windows, and then everything crashed. Consistently, the PSU turned off every time windows was about to start. After a look into the specs, I saw that there was a minimum output for the 5V and 3.3V rails of 0.5A. Because we're only using the 5VSB and the two 12V rails here and the former is seperate from 5V, the load on those rails was exactly 0A. The minimum output for the 12V1 rail is 1A and that seems to explain the weird behaviour.

Apparently, the PC gets into BIOS fine because it needs less than 1A to do so and the PSU does not expect everything to turn on immediately. Somehow that also causes the fan of the PSU, which I suspect to be running of 12V like the one in my main rig does, to turn off. As windows is trying to start, the load gets higher than 1A, and now the PSU has to enforce the minimum outputs of the other rails, which it can't because there's nothing connected to them.
The ingenious solution to this is of course to get an unused HDD and connect that to the PSU. And suddenly, everything works out fine.
thumb.gif




Stage 4: Adding the GPU

So now comes the important question: How will the system behave when we add a dedicated GPU? First, we'll appreciate the ingenuity of Gigabyte to use an open ended PCIex4 header on this board:



This enabled me to just take the GALAX GTX 970 from my main rig and put it straight in there. Then plug in the two PEG connectors you already saw on the picture above, turn the thing by 90° so the fans aren't obstructed, and then turn the whole thing on.



This worked out pretty well.
cheers.gif

Even better news: even when the GPU gets power from the PSU, it will not turn on until the mainboard says so, as opposed to the HDD, which starts spinning immediately and won't stop until the PSU is turned off completely. This means that nobody will have to deal with any timing issues whatsoever when using this setup, but that HDDs and probably SSDs will have to be powered by the mainboard unless you don't need the sleep power modes.

After talking to a rep from Fortron, a distributor of FSP, I found out that the problem of the 5V and 3.3V rails being under the minimum output limit can easily be solved by using a resistor between the rail and ground, or, as I did it, with an LED on the 5V rail. The 3.3V limit doesn't seem to be enforced by the PSU.

I'd say my work is done here, but I should probably make a schematic drawing and comment and upload the code I wrote. The latter is pretty solid, it accounts for all sorts of failures and temporary disconnects. Also I should do tests under load to see how the whole setup holds up to the stress. But for now, I'm just happy to know that this works.
 

·
Registered
Joined
·
103 Posts
This looks very promising, how was the noise from the PSU fan in relation to the rest of the system?

Also did you have the motherboard supply 12V through the PCIe slot? or direct from the PSU like I had done with my build?
 

·
Registered
Joined
·
1,190 Posts
Discussion Starter · #9 ·
Quote:
Originally Posted by QinX View Post

This looks very promising, how was the noise from the PSU fan in relation to the rest of the system?

Also did you have the motherboard supply 12V through the PCIe slot? or direct from the PSU like I had done with my build?
The fan of the PSU was about as loud as the one from the CPU and the GPU, so not noisy at all, but as I said, I didn't put the system under load. Still, I would call that a very acceptable idle noise profile.

I didn't do anything special with the riser, so the GPU was supplied with 12V through the PCIe slot, but less than the 75W an x16 slot can provide. It may be possible that the GPU won't be able to sustain operation when under load because of that, but as it can still draw the full 150W through the two PEG connectors, I somewhat doubt it.
 

·
Registered
Joined
·
1,190 Posts
Discussion Starter · #10 ·
Small Update!

I did a bit of tidying up. Sorted all cables and tucked most of the unneeded ones into void space inside the PSU. Of course they were all properly isolated.



I left two 5V wires, one 3.3V wire, the 3.3V sense line and the SATA power wires alone, but those will be cut eventually. One 3.3 wire has to be connected to the sense line, else the PSU may turn off when put under load, but I can do that internally as well. The SATA power is there because it carries all three voltages, so I can easily measure from there if I ever need to or connect and power any device I want to.

Still, I cut about ten ground, six 5V, three 3.3V, two 12V wires and the -12V wire, and the results look pretty nice and a lot less cluttered than previously with that dumb 24pin ATX connector.



Eight wires for the PC plus four for the GPU? Hell yeah!

I'll also update the OP with a collection of what you need to do in order to make this work.
 
1 - 10 of 10 Posts
Top