Disabling Under/over Voltage Protection On ATX Power Supplies


[C] just recently put together a RepRap. Not wanting to spend the money on a dedicated power supply, he looked around for a cheaper solution and found one in an off-the-shelf ATX computer power supply. These ATX supplies are actually a little finicky when not used in a computer, as [C] found, with voltage drops on the +12 line even when a load is connected to the supply. Undeterred, [C] eventually solved this problem by cutting some traces and grounding a few pins on the protection circuit.

The ATX supply [C] used could supply 25 amps on the 12 volt rail, more than enough for a simple RepRap. There was only one problem: the supply would randomly shut itself off, ruining the print. After a little googling, [C] found some people powering 12 volt amplifiers that were running into the same problem. Their solution was to ground a few pins on the protection circuit. Their supply wasn’t quite like [C]’s so he had to do a little experimentations.

It took a few iterations to get right, but [C] managed to figure out exactly which pins on the “power supply supervisor” IC must be grounded to disable the undervoltage protection. With these pins grounded, the protection circuit of the supply is completely disabled, giving him and uninterrupted 25 amps at 12 volts. If you’re looking for a cheap source of power, it would be hard to go wrong with [C]’s tutorial and his power supply of choice.

64 thoughts on “Disabling Under/over Voltage Protection On ATX Power Supplies

    1. The supplies tend to crap themselves on very low loads. Although it’d require a new power supply rather than just opening the innards of your existing one, if you are heading out to get one then you should find much fewer problems if you pick one that’s known to be compatible with the new Haswells from Intel, as they require the ability to run with extremely low loads. A quick google search for “Haswell PSUs” should provide you a nice long list of power supplies to choose from that should give you less grief, some of which are quite cheap. Personally I’d recommend one of the CX/TX Corsairs as they’re single rail designs so you get everything loaded on that one rail for even less of a chance of tripping the over/undervoltage breakers. I realize it might not be in the spirit of hackaday to specifically avoid having to tinker with the power supply, but again, if you’re already heading out to buy a brick for your bench then you might avoid having to rip it open and void the warranty

  1. Or you can do it the easy way, to disable protection rapidly switch the power on wire for 5-10 seconds and then it wont kick off unless there is a direct short. I have used this method to run 400 watt car amps off of 260 watt PSUs. I would not go more then 1/4 over the max output or the voltage drops.

  2. Wow, the reviews on the power supply he used are abysmal. The price is right, but the PS quality sucks.

    Hmm… maybe a project here? Take a crappy PS, analyze the crap out of it, replace the necessary parts, and get a rock-solid PS at the end… Anyone with enough PS experience not to kill themselves in the process up for doing that? Anyone? Anyone? Bueller?

    1. I’ve actually stepped up to the challenge. My ham radio club had a surplus 6m transceiver that wanted some ungodly amount of amps at 13.8V. 13.8V/20A supplies are out there, if you want to pay the 3-digit price. So, however, are 350W AT supploes, and they’re pretty much free. 15V/20A works out to 300W, so I thought I’d give it a try.

      I ended up stripping out most of the components on the secondary, then reusing the high-current ones. The transformer secondaries were, IIRC, wound with pretty heavy gauge wire, so I ended up paralleling the +/- 12 windings and using the +5 filtering components. Some tweaks to the voltage sense resistor divider and the overvoltage and overcurrent protection circuits, and, after load testing for a day, we were good to go. It’s still running — with overvoltage and overcurrent protection!

      I actually had three supplies, all slightly different, and a number of schematics, some off the web, and some were manufacturer’s application schematics for the control IC (TL494?) they all used. Between those documents and a whole weekend of tracing etches on the PCBs, I learned an awful lot about how these beasts are designed.

      Would I do it again? Sure! It was fun.

        1. I always hear Amateur Radio operators ragging on switch mode power supplies. I have personally used my IC-746 to make numerous contacts on nearly every HF band all running off a Switch Mode PS. It’s never given me any problems. In fact, I don’t believe I’ve ever run into a serious issue related to RF interferance from a switch mode PS. Maybe my equipment just has good fliltering???

          1. Precisely. Ham here too — switchmodes are fine. Most of the grumpiness is from memories of 1980s switchers and expecting something from Harbor Freight to sound great on their rig. Proper power supply construction and filtering makes it a non-issue, as any respectable rig has some PSRR too. Problems can be easily fixed with more filtering.

          2. I was referring to homebrew switchers based around old computer PSUs. The high-quality switchers put out specifically for radio use seem to work quite well.

          1. Yup. I found ATX mod work with 50MHz band but not 14MHz. Probraly because 14MHz is low enough for PSU low-pass-filter. When I transmit on 14MHz band, voltage increased to 15V+ and luckily I place a over-voltage protection. That’s is, feed 12V through a 500mA fuse to drive a relay with a Zener diode. Fuse brusted as voltage was over the limit and the relay cut-off anything behind it.

    2. You can get open frame 12V supplies and not having to go through this
      trouble. I have a whole box of 12V 10A supplies.

      How much current does the RepRap really need?

      As for modding PSU. I have done a few.
      – Modded a wall brick +12V/+5V into a 18V charger for a Thinkpad. Wired
      the 12V and 5V winding in series, added current limit via feedback for the
      – Modded old Mac PSU into ATX. Added On/Off control to the PWM chip, +5V
      standby (with a transplant from a dead supply), +3.3V using a TI DC/Dc

  3. Yeah, there’s usually a reason people put supervisor circuits in their products. If it’s tripping, then some voltage is collapsing. If a voltage collapses, that’s usually because the voltage source can’t provide that much voltage at the current rating you’re asking for.

    By operating the PSU beyond its specifications, you are at best shortening the lifespan of the power supply. At worse, you’re creating a system with unknown side effects.

    1. The guy has started it can provide 25Amps on 12V rail. That’s more than enough clarification. The problem with using a computer power supply loaded at only the 12V rail on that the other rails are not loaded. This they are getting higher voltage (since the PWM is the same for all rails on the transformer). That activates the overvoltage protection on some rail (s), which limits the PWM, so they can stay within +/-5%. This greatly limits the 12V current sourcing capability – the rail becomes current source since effectively it cannot maintain low enough output resistance anymore. This way it starts to drop the output voltage. Since the voltage drops with the load now (even we are drawing less Satan the half rail power) the more load on the rail the more voltage drop. At some point we are ready to hit the undervoltage protection on the 12 volts, which usually disables the output at least for a moment and here we go – we have the strange artefacts on the output without even the half of the nominal load. Other than that there are (usually reputable brand) power supplies, which will sell disables if the power good signal fails during operation. Which comes from the supervisor, which gets fooled again by the lack of load.
      The super visor functions are usually limited to providing power good and maintaining the output within +/- 5% range. The second is usually even implanted even without the supervisor circuit, using directly the feedback loop of the PWM controller chip. Furthermore – the overload/short circuit protection usually operates with separate comparator or greatly increases the dead time control, effectively at least limiting the power output of not completely stopping the entire supply.
      It is really important for electronics requiring +/-5+ stability and really not important for a 150W heater… +3-5 super motors which will have a driver dinamcally maintaining motor current and most likely having power supply range of 24-48 volts.

  4. My first thought was that this supply is probably constructed throughout with the same level of care that went into the printed “WIRNING” visible in the picture. The Newegg reviews seem to confirm that. Lots of failures, some which took the motherboard out with it.

    It’s interesting this can be done, but I sure wouldn’t do it. I’d spring for a better power supply – rather than making a highly fallible supply more catastrophically fallible, then connecting it to something I cared about.

    Further, if all that’s needed is a 12V rail, I wouldn’t bother with an ATX supply unless it was already on hand. If you’re going to buy something, there are plenty of dedicated 12V supplies to be had with adequate amperage ratings, for about the same price. And which don’t require load resistors on another rail to function.

  5. In other news, [C]’s fire insurance has been cancelled since this article was published.

    “uninterrupted 25 amps at 12 volts.”

    “it would be hard to go wrong”

    I think it would be VERY EASY for things to go VERY VERY wrong!

      1. You’re not planning on running your stepper motors off of 12 volts are you? I’m sorry to be the bearer of bad tidings if you are but running stepper motors on 12 volts is pretty unacceptable with your intended application. The why of that has to do with inductive dwell timings but I won’t get into all of that here. Suffice to say you’re going to need a higher voltage power supply. It really is as simple as that.

        Either that or your machine is going to run very slowly with inadequate power at its highest speed. Your target voltage is pretty much the top rating of whatever stepper drivers you’re using is. That is where you’ll get the best performance. That whole operates between 12-35V or whatever your drivers are rated for is bunk. If it says that you want 35V, 12 isn’t going to cut it. Not even really a little bit. OK maybe a little bit, but I wouldn’t accept it.

        If you bought the motors, and the drives, the PSU is no place to cheap out. You’ll only be cheating yourself. You should have bought a toroidal transformer and built a PSU with it. That is the ideal stepper PSU setup. But if you do that be aware of RMS vs peak to peak and what that means in regards to AC transformers and DC rectifiers and filters.

        It is very easy to go over your intended voltage if you don’t take that into account. Good luck.

        1. Many drivers these days are constant current chopper drivers to keep from blowing your stepper coils apart, 12v will do just fine in many applications, This is 3d printing not industrial automation.

          Magnetic fields are a function of current through a wire, so please explain why the voltage should matter if we’re running constant current drivers we can adjust. Sure you will reach a point where your maximum voltage dictates how much current will flow into the coils via inductive resistance and coil resistance, but this happens so fast I severely doubt you will ever see a difference.

          1. Or to put it another way starting with a low voltage you’re defeating the purpose of using a constant current driver. Because the current is going to drop off eventually as you go faster. But the higher the voltage you run the faster you can run before the current starts to drop off. No current of course means no go.

            But 3D printers are inherently kind of low performance compared to other machines that use stepper motors so maybe they are more tolerant of lousy performance?

        2. 3d printing as most of us here know it pretty much started with the RepRap.* Old PC PSUs have always been the standard power supply for those with the 12V rail powering the motors and heaters while the 5V rail powers the electronics. Those who really get into it eventually swap their PC PSU out, usually for a 24V unit to get better performance but the practice of using 12V has been very much proven viable.

          The current home/hackerspace accessible comercially built printers are mostly based on the RepRap experience if not directly the RepRap design. I wouldn’t be surprised if most of them have settled on using 24V but I know I have used early MakerBots that still contained computer PSUs and no more than 12V.

          * – Yes, big industrial printers were around long before RepRap but you had to work in a big corporate or university lab to ever see one. If you were so lucky, congrats. Most of us were not. Even smaller universities couldn’t afford them. Yes, I believe those monsters probably did not use 12V.

  6. it is turning off because he used SHIT for PSU

    >could supply 25 amps on the 12 volt rail

    lol NO, if it could voltage wouldnt sag and it wouldnt turn itself off. Basically he took equivalent of shitty chinese $1 phone charger and disabled only protection circuit there was in it just so he could over stress it.

    Fire hazard.

        1. Output sagging is an indication that it is not well regulated in the
          first place.

          The “regulated” part relies on a certain proportional of load to be on
          +5V. Without the load, the +5V rail feedback path would dominate and you
          are left with no/poor regulations as it violates the PSU design assumption.

          1. I can’t be entirely sure but I thought that was what this article addressed. Decoupling the 12 volt rail from the rest of the supply? I’m not the world’s biggest fan of ATX conversions if truth be told.

  7. There is an easier way. Many of those supplies need a load on the 5v rail in order to work properly. Add a 120 ohm power resistor across the 5v rail (mounting it into the fan’s airstream for cooling) and the supply will crank out 12v all day long. A few large filter caps on the output help clean it up a bit, also.
    I’ve done this several times with power supplies of various quality and was successful each time.

  8. From the production description

    “The power supply boast its design with dual 820uF capacities and other necessary components such as resistors of all kinds, which make it a perfect choice for entry level modders and gaming users.”

    What the fuck is that? Capacities? Resistors of all kinds? It claims to be a 550 watt power supply but if you add up the wattage of all the rails, it works out to over 600 watts. Furthermore the sticker on the side says

    “Max Combined Wattage
    3.3V & 5V = 200W
    3.3V & 5V & 12V = 400W
    Total Average Output 400W
    Total Max Output 550W”

    So presumably it is only recommended that it will put out 16A on the 12V rail, and that extra 9A is only for surge currents such as when the hard drives initially spin up.

      1. ehm, “same price” ? What are you on about? 20USD != 50EUR.
        The one you linked is actually 3 times the price of the ATX.

        Not saying it not still a better buy, but it is certainly not the same price.

  9. another warning for those wanting to hack power supplies is that they can hold dangerous voltages so make sure all capacitors are discharged before messing with the power supplies.

    and test all heat sinks to make sure there is no voltages on them either before touching them.

    if the power supply has been running recently make sure to allow time for the heat sinks to cool before touching them especially on the cheap chinese power supplies since they can put out more heat or are lower rated power supplies marked as higher rated power supplies.

    like taking a 200 watt power supply and calling it a 500 watt power supply.

    it may be possible to use external capacitors to stabilize the outputs and diodes to prevent surges from going back into the power supply

    it is like connecting a 12 volt battery to the 5 volt rail that would also cause it to shut down if you have not burned anything on the 5 volt rail first

    1. I don’t worry about all capacitors, just the high side filter capacitors. If the PSU is wonky they might hold a dangerous charge. If everything is as it should be there is virtually no worrisome hazard in an unplugged ATX PSU though. now *if* is a big word so some caution is still advised. But if a hazardous voltage is discovered in an unplugged ATX PSU then further investigation is warranted. As it really shouldn’t be there, and is easily fixed too. Just put a bleeder resistor across the offending capacitor.

  10. That’s not the right way to do it, you should put 3 trimpots from the 12V line to pins 1, 2 and 3 and adjust them to 3.3, 5 and 12V, that way you keep the protection circuit, and in case your machine consumes more current, the pwm signal will act accordingly and increase the suppy on the primary of the transformer.
    The way you did it, you are overloading the PSU since when the machine sucks current, the voltage drops because the pulse width stays the same on the primary. The normal behavior whould be that when the machine wants more amps, the voltage drops, the pulse width becomes larger and then the voltage stabilizes.

      1. Hey, you’re right. I previosly looked at the datasheet at work and didn’t pay much attention, the feedback loop is at pin #16, pins #1-4 are just over and under voltage protections. But I still think that he should use the protection on pin #3. If it has spikes on the 12V rail he could put an inductor in series, or maybe a small capacitor from pin 3 to ground, It would absorb fast spikes because is charging slowly through a resistor, yet be fast enough to let the protection circuit act.

  11. ATX supplies are not what people should look for if they want high current outputs , instead look for surplus AT supplies. The designs are simpler making them much easier to modify if needed.

    1. You are totally wrong.
      ATX ps differs form AT supply only by a soft start circuit. The rest is the same.
      Different ATX standards have different emphasis, older ones put most of the power on 5V line, nowadays its almost all on 12V line.
      + AT psu will be 200W max, using low efficiency parts, and with dried out caps

  12. I wouldn’t , The caps are rated volts and uf once drained the glow plug affect will happen and direct current is not filtered anymore bye bye motherboard , Just buy a really good power supply and move on…

Leave a Reply

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