Power Supply Running the ATX 12V rail with 18V

scope

Chassis Packer
Original poster
Aug 6, 2019
19
1
Suppose I have a nice 18V output industrial/medical PSU. (But no worries this is for home.) And suppose I'm too cheap to buy that nice HDPLEX 200W 16V-24V input DC-ATX module, but on the other hand I'd also rather not have even the small amount of extra heat from the step-down conversion of 18V->12V inside the case. Let's say I have a 140W long term max. target. At 95% efficiency, that would mean 7W loss; or in a maybe more realistic 80W usage scenario that would be 4W.

And the mainboard already does step-down conversion from 12V with beefy multi-phase converters; so maybe 18V instead of 12V is not that much of a difference in practice?
If I get basic buck converter theory, the inductor ripple current would be just slightly higher when increasing the voltage 12V->18V, and the diode forward current would be like 0.5A larger in this case. So this doesn't look like something the mainboard's 12V step-down converter couldn't cope with easily, does it?

In this case a simpler DC-ATX module would be used, that simply passes through its input onto the 12V ATX rail, and does DC-DC conversion for the other ATX voltages. The module would need to be able to cope with 18V as well, of course. It would be also required that the mainboard's OVP, if it has one, doesn't trigger at 18V. For electrolytic caps 16V and 25V are standard voltage ratings, so better be 25V caps used.

I wonder what else I'm missing?
 
Last edited:

scope

Chassis Packer
Original poster
Aug 6, 2019
19
1
Hmm, PCIe card connector has 12V supply pins, that applies to M.2 as well, so that could be potentially problematic for an SSD.
 

Valantar

Shrink Ray Wielder
Jan 20, 2018
2,201
2,225
While I don't have the level of technical competency to actually say this, there's no way I would see this ending in anything but dead components. If this was this easy, we wouldn't have ATX PSUs fighting to have the most steady and precise voltage outputs.
 

scope

Chassis Packer
Original poster
Aug 6, 2019
19
1
Oh yeah, I forgot about the fans in the bigger picture. CPU fans are normally 12V and I don't think they would like that much overvoltage. (And this may apply to the fan control circuitry on the mainboard as well?)

Maybe PCIe cards exist that step down 18V like they don't care, but PCIe is probably much more critical than mainboards, and this all assuming there's no overvoltage protection anywhere to intervene. In conclusion, there doesn't seem to be much in this.

Now I wonder which input voltage, 18V or 24V, would allow for a more efficient operation for the HDPLEX 200W DC-ATX module.
 

scope

Chassis Packer
Original poster
Aug 6, 2019
19
1
CPU fans are normally 12V and I don't think they would like that much overvoltage. (And this may apply to the fan control circuitry on the mainboard as well?)

Or the controller in the fan itself. I gues something like a Nocuta motor might be able to take significantly higher voltage:
https://www.overclock.net/forum/134-cooling-experiments/134531-fan-24v-overvolt.html
But the PWM IC in the fan must be able to handle 18V supply voltage.

Hmm, PCIe card connector has 12V supply pins, that applies to M.2 as well, so that could be potentially problematic for an SSD.

M.2 has 3.3V supply, so that would be regulated by the DC-ATX module.

But not so surprisingly, OVP on mainboards is a thing, various manufacturers seem to have it under different feature names.