Sounds like a useful project. Although we already have solutions such as the PicoPSU, in some cases, something heavier might be needed.

Definitely following this thread. Good luck!

digger wrote on 2026-10-22, 10:38:

Sounds like a useful project. Although we already have solutions such as the PicoPSU, in some cases, something heavier might be needed.

Definitely following this thread. Good luck!

Thank you!

So the PicoPSU is kind of my inspiration, showing what's possible.

The idea here is to make it so you can actually "fix" the PSU in your vintage computer and preserve it's functional and aesthetic state-- you'd have the same input, same switches, same everything. No DC bricks, no loose wires hanging out everywhere, etc.

This would essentially make it so you can "re-manufacture" old PSUs by swapping out the guts.

bitslasher wrote on 2026-10-22, 10:09:

... I thought a cool name for this project could be the "Revenant PSU" -- revenant being a spirit that reanimates a dead body, so that it can fulfil a purpose or finish unresolved business ...

bitslasher wrote on 2026-10-22, 10:45:

... The idea here is to make it so you can actually "fix" the PSU in your vintage computer and preserve it's functional and aesthetic state-- you'd have the same input, same switches, same everything. No DC bricks, no loose wires hanging out everywhere, etc.

This would essentially make it so you can "re-manufacture" old PSUs by swapping out the guts.

^ Sounds extremely cool to me. 👍

.

I'm gonna float the idea that came to my mind while reading your post: A "+5V strong" ATX "set of PSU guts".

We could fit it into an old (with tired capacitors and outdated, inefficient topology) 80mm-fan power supply. This would keep the classic looks of the PSU.

Or, alternatively we could fit it into some cheap modern-ish (but "+5V weak") 120mm-fan PSU. No classic looks but improved case ventilation.

With a reasonably modern topology and fresh, decent components, for long-term reliability. And while we're at it, with -12V & -5V rails.

I suspect such a thing could be of interest to the retro-PC community ...

TELVM wrote on 2026-10-24, 19:19:

^ Sounds extremely cool to me. 👍

Thanks!

TELVM wrote on 2026-10-24, 19:19:

I'm gonna float the idea that came to my mind while reading your post: A "+5V strong" ATX "set of PSU guts". […]
Show full quote

I'm gonna float the idea that came to my mind while reading your post: A "+5V strong" ATX "set of PSU guts".

We could fit it into an old (with tired capacitors and outdated, inefficient topology) 80mm-fan power supply. This would keep the classic looks of the PSU.

Or, alternatively we could fit it into some cheap modern-ish (but "+5V weak") 120mm-fan PSU. No classic looks but improved case ventilation.

With a reasonably modern topology and fresh, decent components, for long-term reliability. And while we're at it, with -12V & -5V rails.

I suspect such a thing could be of interest to the retro-PC community ...

The last part of your post is a little confusing. It almost sounds like something an AI would respond to a prompt about "A "+5V strong" ATX "set of PSU guts" -- what does that mean?

This project will completely replace all the electronics of the old PSU. No "tired" electronics would remain.

It’s overall an interesting concept per se. There are some things that immediately jump into my mind when reading through it.

The overall topology appears to rely on finding a good, already done 12V PSU (AC/DC). Depending on how much power you need for the system, it might be already a tough task to find one that fits well.
I suppose whenever you design a backplane for your “dead PSU” form factor (speaking of non ATX/AT) you’ll might have to jump for a different base PSU.

This PSU choice is also depending on the amount of power you’ll need. What amounts are we talking about overall?

Looking at what's available, I'd go with one of the larger brands, like the MeanWell EPP-200-12 - 200W, 12V with PFC and <0.5W standby

From a design perspective, I’d make the backplane get the 12V directly from the 12V PSU, as that voltage has to go to the connectors anyways eventually.
The DC/DC part(s) could get the 12V via their backplane connector. Similar how it’s done in modern DC/DC based PSUs anyway. This way the only interface that has to match it the backplane connector + the form factor of the DC/DC module.

If you plan on using the PSU powered by 12V directly, you can just omit the AC/DC supply and connect a 12V jack to the backplane.

Then, coming to what TELVM mentioned, I’d keep in mind that people might need different configurations.

The most demanding rails will always be 12V, 5V and 3.3V. 12V being whatever the AD/DC part offers, and 5V + 3.3V depending on whatever the PSU needs to feed. Socket A for example loves to eat 5V. There could be backplanes with more than one DC/DC socket, allowing for interesting configurations on the low voltage rails:

- single combined 5V + 3.3V module, that provides the usual 20A each
- single 5V module with 40A output
- single 3.3V module with 40A output

(The single voltage modules could be built the same and just be configurable for the right output voltage).

For the negative rails, I’d just slam then on the backplate. The current needed is tiny and you could just have it on that backplate at all time. Give 0.5A or 1A for each and be good. If not, one could have a small module instead that provides the native rails and can be put in optionally.

Then we have the question of soft vs hard power switch.

If you replace pre-ATX era stuff, you’ll probably want to go with hard-power-switching like in the original. Otherwise you’d have to provide a good replacement for the power button in the case. That means additional wiring efforts. Either directly between the AC/DC module or the backplane could have a section to manage the AC wiring as well and expose something like faston connectors to easily connect everything together. (Please no screw terminals).

For soft power you’ll be able to choose two approaches. Either you use a 12V supply with a very low standby power consumption, have it always on and then switch the 12V going to the connectors and DC/DC (just disabling the DC/DC will not suffice, 12V will still need to be switched). Or you have a secondary small standby power supply and switch the 12V PSU’s AC voltage.

For ATX style replacements you’ll need 5V standby anyways. So either do it via 5V DCDC from the 12V rail or the secondary stand by PSU.

Happy to jump in on discussions, like the topic a lot. But being universal always comes at a cost.

bitslasher wrote on 2026-10-26, 05:42:

The last part of your post is a little confusing. It almost sounds like something an AI would respond to a prompt about "A "+5V strong" ATX "set of PSU guts" -- what does that mean?

He basically appears to think about designing a ATX power supply board that is strong on the 5V rail to fit into existing modern ATX PSU housings. To get good 5V performance with reliable components, rather than refurbishing old PSUs.

bitslasher wrote on 2026-10-26, 05:42:

... The last part of your post is a little confusing. It almost sounds like something an AI would respond to a prompt about "A "+5V strong" ATX "set of PSU guts" -- what does that mean?

This project will completely replace all the electronics of the old PSU. No "tired" electronics would remain.

🤣 Good ole human here, no AI shit.

Until the introduction of the Pentium 4 and ATX12V (aka "P4" connector), PCs drew the lion's share of their power from the +5V rail, with the +12V remaining a relatively minor rail. Hence early ATX PSUs up to that time were "+5V strong / +12V weak" (pedestrian terms just to quickly get the idea).

Post Pentium 4 the +12V rail gained protagonism, at the expense of the +5V rail, so modern~ish ATX PSUs are "+5V weak / +12V strong". Tipically 20~25A max on the +5V rail. That may not cut the mustard for some "+5V hungry" classic PCs, like early dual CPU or slot-A Athlon systems.

To power those kind of systems, enthusiasts have two options:

· Keep using a period correct (say year 2000 vintage) "+5V strong" old ATX PSU. With an obsolete topology and (after 25+ years) tired componets, and unreliable, probably faulty electrolytic caps.

· Get a new ATX PSU. With modern topology and fresh caps, but "+5V weak". Risking that such a PSU may be defeated by "+5V hungry" classic systems.

Can your PSU handle Athlon or Duron Socket A processors?

And that's why and where a "+5V strong set of guts" could be of interest 😀

I've also considered the idea of a DC-DC ATX power supply like this, but I think the main challenge is deciding how "modern" the components and configuration should be.
While it is easy to incorporate the latest parts from manufacturers like Analog Devices, TI, or XLSEMI, one must consider whether these components will still be available to an individual—or if they might reach their "End of Life" (EOL) status—should the need arise to repair or rebuild the unit years down the line.
Conversely, opting for "mature" (long-established) components like the TL494 would likely result in lower efficiency. The question is where to draw the line; personally, I would prefer to use mature components, even if it means sacrificing some efficiency.

If I was being lazy, high amp LED lighting PSU kicking out 12V at 50-60A, then those 36A settable modules for 5V and clamped 12V straight through and maybe abuse a serial port chip to get a few milliamps at -12V and zenered to -5v, which will work for most boards that just need to "see" it not use it (watch those soundcard amps) but a better negative supply could be arranged. Though really, one of the cheapest pico PSUs might just be fine for the negatives, and ignore it's weakass 5A or whatever unless you're running a mini 386sx board and use settable high amp module for real 5V ooomph.

Somehow I am lazier still though and always scrape up something useable so never get round to it.

I don't know if anyone has said it yet, but

"USB PD 3.1 in 2021 raised the maximum to 240 W "

For a fair majority of older systems, that is plenty.
It will also greatly reduce initial development time.

All you have to do is accept 48v, and convert it.
12v and 5v make up the bulk of that.
One possible method is to convert all 240w to 12v and then take 5 from that. That makes it relatively simple. 5v can be as high as you want within the 240w total.

Some current sensing and safety cutout would be necessary for the idiots who try stupid things. I'd suggest limiting it to 220w.

Well, using PD3.1 or anything in general begs the question: What level of DIY are we talking about?

A ready to manufacture project/pcb at pcbway/jlcpcb that get's assembled there? Or full diy SMT board population and buildup.
I think this will be the biggest question under all of this - what level of diy'ness is wanted. Same with the choice of components as mentioned by kagura1050 - how much vendor lock in do you want to have and how complicated parts do you want to use for that?

And is highest efficiency really necessary for the implementation itself? Does it matter if the DC/DC section operates at lackluster 82% vs. 96% if using high frequency + expensive parts, if all we're talking about are retro machines.

It's a delicate balance, that will also come into play with cost.

This is why I'd start with exactly defining the requirements to narrow down what "this" project is trying to achieve. Then I'd check potential routes like full diy incl. AC/DC section, off the shelf brick with 12V or PD3.1 to provide main power and then step down.

kagura1050 wrote on Today, 01:17:

personally, I would prefer to use mature components, even if it means sacrificing some efficiency.

Personally, I'd agree on that for that purpose. Doesn't mean we're going back to old grams TL494, but the general premise sounds right to me. Don't go overboard, perhaps with alternate footprints to allow the use of a different regulator population (similar to what's been done on graphics cards for ages).

Pd3 takes half the effort and price away.
People are more likely to buy a dual-use Pd3 charger AND a multipsu than they are too buy a standalone multipsu.

Let's say they cost $10 each
Dedicated = $20 gamble
Pd3 =$10 gamble, and $10 thing you'll probably use for something else if not this.

Human logic isn't perfect but it's not entirely unpredictable.

myne wrote on Today, 07:59:

Pd3 takes half the effort and price away. People are more likely to buy a dual-use Pd3 charger AND a multipsu than they are too […]
Show full quote

Pd3 takes half the effort and price away.
People are more likely to buy a dual-use Pd3 charger AND a multipsu than they are too buy a standalone multipsu.

Let's say they cost $10 each
Dedicated = $20 gamble
Pd3 =$10 gamble, and $10 thing you'll probably use for something else if not this.

Human logic isn't perfect but it's not entirely unpredictable.

I agree on the premise of usefulness, but not on price.
PD3.1 with 240W is pretty scarce, at least from what I see, in terms of adoption. A PSU I can find costs around 90€ - useful, but not cheap and has a single output
PD3.0 with 100W has higher adoption, but it'll limit the power you can draw from it. Bringing me back to the question above: What is it we're trying to power? What power requirements are there. That needs to be fixed before deciding on a source. Or if you say it has to work with PD3.0 (20V 5A) - you'll have to limit the use-cases.

Also to note: PD100W and above need special cables. It's not simply USB-C to USB-C and you're good.

If I can prevent somebody having to assemble anything with mains voltage, I'd be happy. But in that case one might wanna go for a plain ol' barrel jack PSU that fit's the purpose + DCDC behind.
That will leave you with a different than original experience: A powerbrick that you did not have before.

So it boils down to: What issue is this trying to fix. Name and specify that and then a solution can be found / built. It using the original mains input mandatory or not? Do we only need <=100W?

Is there anything stopping it from scaling?
Need 100w use 100w.
Need 240, use 240.
The bigger ones are likely to get more common over time.
Microcontroller will know what the max is. They're common enough.
Essentially, you'd just be copying a laptops internal psu.
Presumably there are boardviews around to help select components.

myne wrote on 37 minutes ago:

Is there anything stopping it from scaling? Need 100w use 100w. Need 240, use 240. The bigger ones are likely to get more common […]
Show full quote

Is there anything stopping it from scaling?
Need 100w use 100w.
Need 240, use 240.
The bigger ones are likely to get more common over time.
Microcontroller will know what the max is. They're common enough.
Essentially, you'd just be copying a laptops internal psu.
Presumably there are boardviews around to help select components.

Not stopping, but it dictates the choice of topology and parts.
If you go 3.1 - you'll have to make everything work with 48V. If you go 3.0 - you'll have to make everything work with 20V.
Again, it's a design choice based on requirements.

Hi, I've been thinking about this issue on-and-off for a couple of years actually, and I just made a first working prototype of custom DC-DC PSU board, I posted this in another topic (AT(X)Anything: a modular DC-DC PSU for old PCs (open hardware)).

Basically it's a supervisor and power distribution board, which can be set up to work in AT or ATX mode, it generates 5VSB and -12V/-5V rails, has multi-stage OVP protection, and for 12V, 5V and 3.3V converters you can put anything you want through screw terminals, or solder smaller modules directly to the PCB. That should mitigate any supply chain issues, because the parts on supervisor board are common and easily available, and it is easy to change what kind of converter you want to use.

I designed it around 20V input, though it also could be scaled further, should work up to 24V with no changes (so USB-C PD 20V or 24V industrial bricks should work).