A Very Tidy ATX Bench PSU

If there’s one thing that for decades of desktop PCs have given us, it’s a seemingly endless supply of relatively capable power supplies. If you need 5 volts or 12 volts at a respectable current they’re extremely useful, so quite a few people have used them as bench power supplies. Some of these builds box up the mess of wires into a set of more useful connectors, but [Joao Pinheiro] has taken his to the next level with a very neat 3D printed case and a set of variable switching regulators to make a variable bench supply with a top voltage of 60 volts.

In many ways it’s a straightforward wiring job to build, but there’s an unexpected power resistor involved. It’s sinking the 5 volt line, and we’re guessing that some current is required here for the PC power supply to run reliably. The thought of a high power resistor dumping heat into a 3D printed case leads us to expect that things might become a little melty though.

ATX power supplies are so numerous as to be expendable, so it’s always worth regarding them as a source of parts as well as a power supply.

20 thoughts on “A Very Tidy ATX Bench PSU

  1. It might be pretty but it’s still cost optimized switching PSU with enough ripple to trigger TTLs when under no load. There’s a reason why electronic boards inside computer have so much capacitors it’s unbeliveable.

    1. You are so utterly mistaken it’s unbelievable. Those capacitors are primarily for DCDC stepdowns to around 1V where going from a few hundreds of mA to 60/80/100A is extremely frequent thing for a CPU (cos of dynamic underclocking). For GPU the sudden consumption spike can be even greater even though less common and those also run about 1.x Volts using stepdowns from 12V.

      Modern entry level nongaming PSUs about 80€ are stable AF and has been for at least one and a half decade. And the 3.3V and 5V rails are superdooper stable because thouse are used directly unlike 12V rail. If you buy 20$ Aliexpress PSU that is the problem.

      1. I call bs on your theory. I in 2008 I bought Fortron FSP 450W PSU and it isn’t stable at all. Replaced capacitors in 2010 but voltage still dropping below ATX spec when playing GTA4, ARMA2 or Red Faction Guerilla. Only Tibia and FlyFF work ok. Unfortunately I could not been able to return it because it was bought on Allegro instead of dedicated store. Since then I try to avoid ATX when possible. My Toshiba L300 laptop uses 80W Mean Well Medical-grade PSU and it’s rock solid compared to my PC with Intel E8400 and Nvidia 8800 GS. 450W should be more than enough to power it, especially after removed DVD burner and replaced 2x 250 GB RAID0 Seagate HDD with single 500 GB Kingston SSD.

      2. You have to load the power supply down a bit. I did this, and when you put a resistor across the 12V rail, things settle down, and the output can be fairly clean. A zero load will cause these power supplys to oscillate and become unstable. At least a 10% of Max load must be added to stabilize things. Adding additional Caps doesn’t hurt either. The resistor has to handle about 10% of the power supply max capacity. This does mean a big heat emitting power resistor. That’s the bummer, but if you can find a place to put this space heater, then these power supplies can work quite well. Never mind the 10% loss in the form of heat.

    2. I bought a pair of 30V, 5A Multicomp linear power supplies from Newark about a year ago. They look suspiciously like they’re actually made by Owon. The cost was a lot less than anything I could make myself and had some neat features such as, 1 mV resolution, SCPI control over RS-232 and a 3.5 inch front panel display which can graph current or voltage output. Inflation has raised the cost almost 50%, but they’re still very affordable and less expensive than anything which can be home brewed.

      The little Multicomp units aren’t as clean on the output has my large Rigol 832A supply, but they’re acceptable for non critical analog bread boarding. My gold standard for noise free linear supplies is an HP dual channel +/- 50V supply/amplifier made in the 1960s. There isn’t a lick of digital anything in sight. The HP isn’t meant for 1 mV output control, but the DC output is rock solid. Even the fans are linear, just shaded pole box frame motor driven.

      Years ago, I needed a constant current source for a tactile force stimulator for lab use. We originally used photo flash batteries which proved very expensive. We were recording surface EEG signals, so any power supply noise was a huge issue. My solution was to build a high power sinusoidal oscillator running at about 1 MHz driving an air core hand made transformer and voltage doubler to get 200 Volts to supply the current source unit. The oscillator was driven by a car battery kept charged off line. There was still noise, so I ordered a wide band 600 ohm : 600 ohm audio transformer meant for use in mixing consoles. I connected one winding in series with each of the output leads such that the common mode noise would be bucked out. That was enough to clean up the supply to get good EEG recordings. There is both a lot of science and art in making a good power supply.

  2. He uses a 50W 10 Ohm resistor on the 5V rails.
    But that dissipates only 500mA and thus 2.5W.
    Those orange metal resistors are however very sensitive to overheating and only meant to be used on a heatsink. Placing them near a fan is no substitute for a heatsink. If you want to use it without a heatsink, then one of the “cemented ceramic” resistors is a much better choice.

    I do like the addition of the SMPS modules and meters to get variable voltages and current limit. An adjustable current limit should be in the list of minimum requirements of any lab power supply. It does not have to be very accurate (for most applications), but if your circuit for example needs 50mA, and you set the current limit to 100mA, then it can save your circuit in case of shorts or other mis-wiring. At such a low current there is often no other damage, while at a higher current different flavours of magic smoke get released.

    A think I do not like though are the many fuses on the front panel combined with the banana plugs. I would move the fuse holders elsewhere so you have more room and can space the banana connectors further apart.

    Another consideration is to spend a little bit more money on the meters. For a modest amount of extra money you can buy panel meters with one or two extra digits.

    1. The 10 Ohm resistor on the +5V line is probably completely useless. Ancient PSUs needed a base load to start and/or regulate properly, but modern PSU should do fine without it.

    2. “… but if your circuit for example needs 50mA, and you set the current limit to 100mA,…”
      then you’re probably gonna want the current limit on the lower voltage output.

    3. Hi Paul, you’re right on the dissipation values, however the thermal dissipation capacity of a 50W resistor such as the one used is quite better than a 5W one, specially ceramic ones.

      I agree with you on the fuses – in the end, the front panel is a bit cramped and could probably benefit from a makeover.

    4. I put the load on the 12V rail. It was a huge 50W power resistor. I didn’t want the resistor to get dangeriously hot so I went big. It turned out to be about right, and that resistor would get real hot. I had to keep the resistor separated from other things because of the heat. The smell was much like the old tube TVs with a bit of burning dust.

  3. Some years ago there would be hacks which hacked the power supply itself to manipulate the reference voltages and things like that, no need for cheapee regulators downstream

  4. The gap around 11V wouldn’t fit well for me. I have found that those LED RGB strips run fine on less than their speced 12V with their controllers. All the way down to 9V. At the 3 cell lithium range which can be less than 12.00V, they will last longer as 12V feels hot at each LED. Trying to find the optimum voltage to run a gadget with a lost power adapter and not fry it is where continuous range is needed on a power supply, with volt and amp meters.

  5. If you use ex computer power supplies to power amateur radio equipment its a good idea to add extra output of filtering as they can generate rf noise but I notice that linear power supplies using 723 voltage regulator m&s cause rf noise and need extra ceramic filtering and bypass capacitors

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.