1000W max server PSU mod - 12v external brick

[timginter]

WARNING: THIS GUIDE DEALS WITH DANGEROUS ELECTRICS! ALWAYS UNPLUG A POWER SUPPLY BEFORE DOING ANYTHING ON IT, ALWAYS PLAN WITH SAFETY IN MIND AND DO NOT OPEN A SERVER POWER SUPPLY!

I needed an external brick able to provide 400W at 12V, with potentially 550W when I upgrade my GPU. I ended up modding a 750W server PSU... and it even works! With 750W max, the 400W peak I currently pull doesn't make the fan spin loud, it's just a low hum, barely audible even at 400W load.

PARTS:

• server PSU, HP 750W in my case, around £15 on ebay
• around 1800mm of 10 AWG cable (e.g. 900mm black, 900mm red)
  • I used https://www.wirebarn.com/Wire-Calculator-_ep_41.html to calculate required wire thickness
• around 2000mm of 18 AWG cable
• on/off rocker switch
• 2 pairs of XT90-anti-spark connectors
  • that's why 1000W max - XT90 is rated for continuous 90A current
• 1 pair of XT60 connectors
• around 1500mm of expandable cable jacket (or 900mm of stiff one)
• shrink wraps
• hot glue gun
• loctite / super glue
• liquid tape
• soldering iron
• snipe-nose pliers

 

[timginter]

TESTING THE PSU

Server PSUs may be different - find how to activate/energise it first (ideally before you buy it). If you can't get past this step, there's no point in going further. Mine is HP 750W server PSU, Gen8 660183-001:


I used the rocker switch and 18 AWG cables to test the PSU - I could test the switch at the same time. Rocker switch had holes in terminals, so I just hooked the cables:


I used cloth pins to more or less safely attach the cables to the PSU. On the model I bought, to activate/energise the PSU you need to short the left-most pin and the shortest pin (on the top):


All set up before the test:


And after the switch:

 

[timginter]

POWER SUPPLY

The trickier part next - soldering connectors to the PSU. XT90 is great for main connectors, XT60 has the perfect gap to solder straight to smaller pins to activate the PSU.

I'm not very comfortable with a male XT90 connector (the one with two metal pegs) on the PSU - just feels too easy to short by accident. Same for the XT60 to activate the PSU - female connectors (with two holes) just feel safer - no accidental activation, no accidental shorts.

I laid down a mock of how things will connect to the PSU:


Then soldered the connectors straight to the PSU:


That was a proper pain... I actually made a mess of the XT60 - I pulled the pin I needed together with old solder. In the end I just used some loctite/super glue to get the connector in place first, then soldered it to the pins. I had to solder to a nearby pin and bridge the missing pin with solder. Then, not to short unwanted pins, I had to scratch the connectors off of the board... proper mess.

I used hot glue around the connectors to keep everything more sturdy and a thick coat of liquid tape to make everything safe:


I've seen turds looking better... :/ I should have smoothed out the hot glue before putting the liquid tape :/ It works, though - I'll call it a prototype.

After having it done this way, though, I actually ordered another HP 750W PSU, same model. I'll try soldering the connectors vertically, pointing up - soldering will be MUCH easier with proper metal-to-metal contact, the PSU should end up shorter, less prone to catching anything with the connectors and just nicer looking in general.

UPDATE:
On the second try, XT90 was really easy to clamp and solder vertically, but I messed up the XT60 - the second PSU makes a squeaky/buzzing sound when activated and I'd rather not use it - I'll stick to the prototype. Vertically soldered connectors looked good, tough, but would also need hot glue to make them more sturdy, especially the XT60.

 

[timginter]

CABLE

Around a 900mm length felt good - my laptop lead is 1200mm from the brick to jack, but I didn't want too much length in case there's a bigger voltage drop at high load.

I wanted a rocker switch next to the female plug which went into GxR-ONE plug-in DC-ATX power supply - easy to activate/deactivate the PSU before I unplug it. I had the cables hooked from the test before and with the holes in terminals it didn't feel like it needed soldering - I just used shrink wrap around to keep it in place:


And added another shrink wrap to keep the two cables together:


And another one to hide the whole switch - I just stretched a medium size shrink wrap with snipe-nose pliers:


I wrapped all PSU-end cables together with sellotape and threaded them through the jacket:


Expandable jacket get really stiff when stretched - that's why around 1500mm for a 900mm cable - it will gather and keep the cable flexible.


Set up and added a bigger shrink wrap which will also hold the jacket in place.


I soldered the connector, twisted the jacket into the shrink wrap and finished this end:


From the layout when soldering the PSU, I started with soldering the male XT90 connector first - that's also the shorter cable:


Each cable has its own shrink wrap + a bigger one to keep both cables together and hold the jacket.
When the male XT90 connector was done, I measured how long the thinner cables should be to easily reach the female XT60 on the PSU - I left extra slack so if I pull the cable, I'll be pulling the PSU by the thicker 10 AWG cables instead of the thin 18 AWG.
Soldered the male XT60, stretched the shrink wrap with snipe-nose pliers to fit over the whole connector and wrapped the jacket:


And the cable finished:

 

[timginter]

FINAL TESTS

After everything was done I tested the PSU while running a benchmark - the longer the cables, the higher the voltage drop will be and if I messed up any soldering voltage drop should highlight it. ATX spec lists +12V rail with 5% tolerance, between 11.40 and 12.60V.

PSU had a stable 12.33V on terminals without load. I used a male-to-male XT90 to connect the PSU with GxR-ONE plug-in ATX PSU - gave me easy access with a multimeter:


With the prototype (XT90 soldered horizontally), my multimeter measured 11.98-12.22V - max 2% drop at around 350W - better than I expected from the wire size calculator. Well within ATX spec and good to use

 

[harveyy_sh]

nice! does it turn off automatically when you turn your pc off? or does it need to be turned off manually?

 

[timginter]

nice! does it turn off automatically when you turn your pc off? or does it need to be turned off manually?

Thanks.
It doesn't turn off automatically, I have to flip the switch to deactivate it and/or unplug.

Also, without any further mods the fan is turning even with the PSU deactivated but plugged in. It's a quiet hum, around 15-20dB, though

 

[CC Ricers]

I guess using a relay of sorts could help turn the PSU automatically off without needing to press another switch? I'm no electronics wizard though, the most complex thing I've done thus far is wire a BMS to batteries.

 

[timginter]

I'm using the PSU like an external brick, with any animals or kids in the house I actually prefer a separate switch - I can leave the PSU connected to my PC but deactivate it separately. The humming when plugged into a socket reminds me to unplug it for extra safety, too.

I was thinking about a 0 rpm mod for the fan (e. g. with a temperature sensor) to stop the humming when PSU is deactivated but plugged in, but it didn't bother me enough to open a 750W PSU and play around with the cables

You could use XT90I plugs - they have an extra pair of smaller pins, could be wired to activate the PSU at the same time you connect it to the PC instead of a separate switch