4
\$\begingroup\$

I am building a miro-itx slimline pc for my mobile home that is powered by solar. I have selected a GPU with a rated TDP of 120 Watts and a CPU with a TDP of 53 Watts.I purchased a DC-DC buck converter ATX power supply the M4-ATX 250w rated for 12A continuous on the 12V rail. I made the incorrect assumption that the CPU was powered by an onboard buck converter working on the +3.3V rail( would make more sense? It's rated @ 25A), and thus the CPU load would count towards the maximum. After a little research it would appear the CPU buck converter draws from the 12V rail, pushing my power demands out of spec. I have thought of a few solutions and am curious as to their viability.

1.) Parallel another power supply onto the 12v Rail (Buck/Boost with Constant Current set to enough amps to bring the unit back into spec Constant Voltage mode set at higher voltage then my current 12V feedback locked DC-DC M4-ATX).

This would effectively make the Device constant current, and lock it to the regulated voltage.

or

2.) Parallel another M4-ATX onto the current one and hope that line regulation is sloppy enough to prevent short circuit-like feedback loop.

I would prefer #1 But I do not want any fried expensive electronics.

Thanks for reading through my word wall.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ Before people START downvoting - please read carefully. This IS a design question. \$\endgroup\$
    – Russell McMahon ♦
    Commented Jul 24, 2015 at 4:51
  • \$\begingroup\$ You could also CC buck from the 3.3V rail as it has lots of capacity, with probably a cutout if voltage rose more than slightly too high. If 3v3 has enough pare capacity you may even be able to linear regulate to CPU supply. \$\endgroup\$
    – Russell McMahon ♦
    Commented Jul 24, 2015 at 4:54

1 Answer 1

Reset to default
3
\$\begingroup\$

You question is about the viability of your two suggested options. Both options are essentially the same thing: load sharing between two power supplies-- in one case it's two of the same supply model, and in the second case it's two different supply models.

Load sharing between two switching converters requires some shared loop control to operate properly. Some power supplies have connections that facilitate ganging of supplies like this. It looks like the M4-ATX does not.

If you try to just tie them together, you will have two supplies each with different internal control loops that respond differently (that is, with different timing) to the downstream load transient conditions. This can result in poor regulation and/or oscillation of the shared 12V output. This could be minor (you don't know it's happening), major (what's that sound?), or catastrophic (what's that smell?).

Your best bet is to get a supply that can support the full load.

\$\endgroup\$
2
  • \$\begingroup\$ @drbrwn I have chosen your advice after attempting to hack it. I attempted to use a simple buck converter soldered into the 12v rail and using measurements tune it to supply ~7 amps additionally on the rail. It worked fine during testing when connected to my Lab PSU but when I switched from 16AWG test leads as the primary conductor to 8AWG the smoke was released from one of the capacitors on the chinese buck converter \$\endgroup\$
    – Shane
    Commented Sep 13, 2015 at 2:18
  • \$\begingroup\$ I think the combination of budget cap's on the ebay buck converter and lower resistance along the input power line isolated the capacitors as the weak link along the current path. The capacitors in question were upstream of the m4-ATX's dc filtering capacitor bank which probably caused a mush high peak current. This is all just my best guess. I am curious if you have any ideas as to why the failure occurred. (again the only variable that changed was the increase from 16AWG to 8AWG). \$\endgroup\$
    – Shane
    Commented Sep 13, 2015 at 2:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.