Open-source PSU breakout boards for the retro bench, from simple ATX breakout to an all-in-one tester

[fargo]

Hi all,

Over the past few weeks I've been designing, with AI help, a small family of open-hardware boards for working with power supplies for vintage machines.

v1.0 - The Breakout Board
ATX-24, AT P8/P9, and disk Molex connectors broken out to banana posts and screw terminals, with per-rail blade fuses and rail LED indicators. One terminal ("XV") doubles as a 9 V AC/DC entry riding the 3.3 V rail, made with Commodore-style bricks in mind.



Example usage:

• load-testing a PSU using external loads connected to the banana posts
• powering an ATX or AT motherboard from a bench power supply
• powering an AT motherboard using an ATX PSU

v1.1 - Dummy Loads
Same breakout as above, plus four TO-220 power resistors on a heatsink bar with the ability to select load profiles. The profiles provide proper testing of old PSUs for PCs and micro-computers such as Commodore and Atari.



Example usage:

• burn-in an AT/ATX supply at ~30 W on the PC profile
• test Amiga/Atari ST bricks on MICRO
• load-test a C64 PSU including its 9 VAC winding

v2.2 - Voltage Generation & Protection
Crowbars on +5/+3.3/+12 V with trip points tight enough to catch a wandering rail before your chips do, plus reverse-polarity shunts on every rail. It can also generate +3.3 V, −12 V, and −5 V from a +12 V feed.


Example usage:

• power an ATX or AT motherboard from a dual bench power supply
• generate the −5 V/−12 V that many modern ATX units dropped
• give a 3.3 V rail to an ATX machine running from an old AT supply that never had one
• put an overvoltage tripwire between a suspect PSU and irreplaceable silicon

v3 - All of the Above on One Board!
Breakout + loads + generators + crowbars, five toggles, one 100 × 198 mm board. Modes compose:

• Direct: acts as the v1.0B board (except it can't take AC input)
• Load-test: acts as the v1.1 board (again excluding AC input)
• Protect: acts as the v2.2 board with crowbars armed
• Generate: acts as the v2.2 board making the missing rails
• …and combinations, e.g. load-test a supply while protected

It also features polarity-swap protection on the screw terminals. The first two v3 test boards are at the fab now.



KiCad sources, schematics, BOMs, and a frank risk analysis are all on GitHub: https://github.com/techana/psu-breakout. Feedback, critique, and build reports very welcome

 

[BloodyCactus]

have you tested your trip points on modern atx psu? being able to dump 70A, 80A to 100A into a 12vdc rail is uh quite scary without good overvoltage/overcurrent protections. (when someone asks me I usually try and dissuade them using an ATX psu for things like bench power supplies).

 

[fargo]

have you tested your trip points on modern atx psu? being able to dump 70A, 80A to 100A into a 12vdc rail is uh quite scary without good overvoltage/overcurrent protections. (when someone asks me I usually try and dissuade them using an ATX psu for things like bench power supplies).

Good question. The board was originally designed to receive input for powering old devices through the banana posts. Connecting a PSU to the ATX connector was meant for testing that PSU, not for powering other devices through it. This means the safest way to use a modern ATX unit as a bench supply is to feed the board via the banana posts as that path gets every layer of protection the board has: the per-rail blade fuses are in the fault loop (5A for 12V by default), plus the precision crowbars and reverse-polarity shunts on the rails.

Powering other devices from a PSU plugged into the ATX connector is a byproduct of the pass-through design, and it does not get the same protection. In that case the connector-to-connector path is unfused, same as inside a real PC. So the board's protection is limited to what the crowbars and clamps can do: an overvoltage still fires the PSU's short-circuit protection (SCR) and clamps the rail, and reverse-polarity faults are still clamped, but interrupting the fault current depends on the PSU's own short-circuit protection, not on a board fuse. With a modern unit that has working SCP, that's a millisecond shutdown; with a vintage or faulty unit that can sustain the current, the SCR clamps at around a volt and becomes sacrificial. RISK_ANALYSIS.md in the project's repo spells this case out.

 

[Plasma]

Have you actually built and tested these or are they just initial designs?

I find the version numbering a bit confusing since it seems like these were all designed at the same time. If you make improvements to the v1.0 board, what will it be called?

I understand you don't want to be liable if someone damages equipment with one of these boards, but the statement "not intended for real, professional, commercial, or safety-critical use" on the github page doesn't instill a lot of confidence. Not intended for real use? Do I pretend to use it?

 

[fargo]

Have you actually built and tested these or are they just initial designs?

I've only built the initial design (named v1.0A in the repo). Two v3 samples are being fabricated now and I'll receive them in the coming weeks. I'll report my findings then.

"I find the version numbering a bit confusing…If you make improvements to the v1.0 board, what will it be called?"

Sorry for that, but this is how the project evolved! I initially wanted a simple board to make PSU testing easier using an external DC electronic load, and I came up with v1. Then I thought, why not put the load on the board instead of using an external tool? So I came up with v1.1. Then I thought: what if I'm using an unreliable power source, or make an unintentional mistake while testing precious equipment? The board should protect against that! So I came up with v2, which I improved into v2.1 and then v2.2. Finally I said "OK, I need to combine everything on a single board!" and v3 came out.

As for what an improvement to v1.0 gets called: that already happened! The repo has v1.0A (the original I built) and v1.0B (the improved revision posted above). Letter suffixes are revisions of a board; new numbers are new capabilities.

"…'not intended for real, professional, commercial, or safety-critical use'…doesn't instill a lot of confidence. Not intended for real use? Do I pretend to use it?"

Fair point! "real use" is clumsy wording on my part, and I'll fix it. What I'm trying to say: this is a hobby project for the hobbyist bench, designed by one person and not certified against any standard. The designs are there for anyone to build on or improve, and whoever fabricates one should understand what they're building. The repo's RISK_ANALYSIS.md exists precisely so that's an informed decision rather than a leap of faith.

 

[Istarian]

If the objective is just to power another board it might be easier to just take a fixed DC voltage from a bench supply (or other source), produce the desired voltages, and route them through a connector.

 

[fargo]

If the objective is just to power another board it might be easier to just take a fixed DC voltage from a bench supply (or other source), produce the desired voltages, and route them through a connector.

Just so you know, any of the boards I mentioned can do that. The tricky part is that most computers need more than just one or two voltages, and getting a multi-channel bench power supply can be expensive. That’s why I created v2.2 and v3 boards, which can produce +3.3V, -5V, and -12V from a +12V input.