I am considering making a 12V UPS for an ATX computer. I would like to know if my concept is sound, (at least halfway so :) )

It's insane that commercial UPS rectify+step-down mains to charge a 12V/24V battery, then invert/step-up this battery voltage back to mains voltage, only for it to be again recified/stepped-down to 12V DC inside the computer PSU. I want something more efficient.


  • Use existing ATX power supplies (only the 12V rail)
  • Use existing salvaged Li-Ion cells (not LiFePO)
  • Super minimalist approach to component count and circuit complexity
  • Maintain high efficiency by avoiding DC-DC conversion losses


To avoid losses, no DC-DC conversion is to take place on 12V. The ATX spec allows ±5% on the 12V rail. Let's be very naughty and stretch that to -10% on the low end, giving us 10.8 to 12.6V. Let's put 3 Li-Ion cells in series (including a suitable BMS).

For longevity and ease of finding a suitable power supply, we shall charge the cells to only 12.0V ≈ 4.0V/cell. On power loss, the cells discharge and under-voltage protection needs to kick in at 10.8V ≈ 3.6V/cell.

The 3.6V-4V range gives us slightly less than half of the cell's energy capacity, but that is sufficient. In turn, the longevity of the cells will increase fourfold. (Source: Battery University's Discharge characteristics and Prolonging articles)


For simplicity's sake, I'd like to avoid having to switch to the battery on power loss (i.e. offline UPS). Instead, the battery shall be permanently paralleled to the power supply (online UPS). Diodes to avoid having the battery back-drive the PSU shouldn't be necessary, as SMPS already have diodes on their outputs.

Charge current limiting

The battery can't exactly be parallel to the PSU, otherwise there would be no way to limit the charge current. Something like this circuit will be necessary:

current limiting


ATX power is more than just 12V. Areas of concern:

3.3V and 5V rails

These are also required. They can either be supplied by a mini DC-DC converter behind the actual PSU/UPS, such as as these Inter-Tech or Mini-Box products (since those rails need very little power on modern systems, the switching losses are negligible)

Or by using a modern ATX12VO or pseudo-12VO mainboard, such as an Asrock Rack one which only uses the 12V rail.


The UPS would need some minimal logic to handle the ATX PS_ON pin correctly. It should monitor PS_ON and switch on/off accordingly, aswell as passing this signal through to the actual PSU.


The PSU signals to the mainboard when voltage rails have stabilized. Obviously this signal will drop when the PSU switches off during power loss, resetting the PC. The quick&dirty solution would be just tying this signal to 5V directly.


Will this work? Apart from being naughty and over-exploiting tolerances, are there any glaring mistakes or omissions?

Thanks a lot!

  • \$\begingroup\$ So you don't want to use DC to DC conversion due to losses... Then you state you will be using DC to Dc conversion to get 3.3V and 5V rails. A start would be a clear position. \$\endgroup\$
    – Solar Mike
    Sep 7, 2022 at 13:46
  • 1
    \$\begingroup\$ You are greatly limiting your energy storage if you're limiting your li-ion cells to 3.6-4V. They usually are a lot happier at 3.2-4.1V, look at the discharge curve, and you'll realise just how much of thier cycle happens below 3.6V. And that's before we look at voltage drop due to current draw. What kind of currents are you planning on? You'll need a lot of li-ion cells. Many SMPS do not like to be back fed, so diodes will be needed to make sure they run well. \$\endgroup\$
    – Puffafish
    Sep 7, 2022 at 13:47
  • \$\begingroup\$ >Then you state you will be using DC to Dc conversion to get 3.3V and 5V rails. -> Yes, but I also state that these rails only draw negligible current. \$\endgroup\$
    – 1N4001
    Sep 7, 2022 at 14:24
  • \$\begingroup\$ @Puffafish Good points, thanks. I do have a lot of salvaged cells. The system will be mostly idle, let's assume 5A total, spread over 4-6 cells in parallel. So 1A or ~0.5C per cell. 1/2 total capacity across 3.6-4V was too optimistic on my part, it'll be closer to 1/3. \$\endgroup\$
    – 1N4001
    Sep 7, 2022 at 14:46

2 Answers 2


It's insane that commercial UPS rectify+step-down mains to charge a 12V/24V battery, then invert/step-up this battery voltage back to mains voltage, only for it to be again recified/stepped-down to 12V DC inside the computer PSU. I want something more efficient.

"Inefficient" and "insane" don't mean the same thing.

My desktop PC is plugged in to a UPS right now. Guess how many minutes this PC has run on battery power in the two years since I bought it... Zero. Zero minutes of inefficient power delivery every two years is not a big problem for me. Meanwhile, it gives me peace of mind to know that my files won't be trashed if the power ever does out while I am working.

Not every PC owner in the world wants to pay for that peace of mind. Go figure! But that's why commodity PCs don't all have UPS built in. Given that my PC doesn't have UPS built in—doesn't even have any provision for an external DC power supply—the UPS box at my feet is a reasonably cheap and easy solution.

Cheap, easy, sane.

  • 1
    \$\begingroup\$ This does not answer the question. Question is not about efficiency or sanity. It's about whether the plan works or not, mistakes and omissions. \$\endgroup\$
    – Justme
    Sep 7, 2022 at 15:43

Won't work.

A power supply is not a lithium charger so it can't be connected directly to a battery.

Lithium batteries need CC/CV charging and charging needs to be stopped when charging current is low enough. They can't be kept floated with charging voltage indefinitely.

If batteries are drained to say 10V and the ATX supply is turned on, it will try to raise voltage to 12V but batteries will draw all current they can and voltage won't rise to 12V. ATX supply likely shuts down due to overcurrent or undervoltage.

Likely the power supply would also go dead with a puff of smoke when back-fed from another source.

For fire safety, don't use lithium batteries improperly with custom hacks.

And you still have to make the other voltages for the system, and not just for the motherboard but all hard drives, add-on cards, and USB devices. Even a basic CPU and video card could be 100+ watts.

Even if it worked, you would be left with computer running, but no display or other mains powered items such as external drives, network switch or router.

At least I would suggest making this on a computer such as NUC with 12V input or similar.

  • \$\begingroup\$ There is charge current limiting, making it CC/CV. Of course the cells can be kept floating, especially at the low final voltage. Why not, they're not NiMH. SMPS have diodes on their output to avoid backfeed. Generating the other voltage rails is alos covered in the proposal. \$\endgroup\$
    – 1N4001
    Sep 7, 2022 at 14:30
  • 1
    \$\begingroup\$ Well, experiment each stage then, if all stages, work then it works. You could start by feeding 12V into an turned-off ATX supply then, but I expect damage. If you only need 60W, getting a laptop could be safer option for a battery backed up computer than risking your motherboard and other components with a custom solution. \$\endgroup\$
    – Justme
    Sep 7, 2022 at 15:21

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.