Adjustable power supply using computer SMPS

Discussion in 'The Projects Forum' started by tom66, Nov 8, 2010.

  1. tom66

    tom66 Thread Starter Senior Member

    Joined:
    May 9, 2009
    Messages:
    2,613
    Location:
    Basingstoke, UK
    I discovered a very cool thing to do with an old computer SMPS. I thought people would want to know.

    In brief, it involves converting the supply so instead of outputting a fixed 12V/5V/3.3V, it outputs a variable 7V-10V/3.5V-4.5V/2.2V-2.8V. I'm working on improving the range so I can get the outputs to go above 12V (with caution, as the filter caps are only 16V rated) and below 2.2V. The voltages are adjusted in harmony using a potentiometer. It is unfortunately not possible to control each rail individually. I used a cheap Elenco 120W power supply, from a very old computer, but any supply should work.

    The first thing you must note is that a computer PSU is mains operated equipment. Therefore, it is easy to get shocked. You must take appropriate precautions including discharging the main filter capacitor(s) before modifications commence. If not, you will get a nasty shock, it could even be fatal. Don't be lazy, and double check the caps with your multimeter!

    The second thing to note is you must have a special type of PSU. It must have a feedback path to the control IC on the secondary side. In my case, the 5V rail was the feedback path. It went through a voltage divider (very important) to a TL431 shunt reference, and to an optoisolator. The TL431's configuration is like a comparator: it conducts when the supply voltage exceeds a set point. This allows it to be used to produce the feedback signal, which, along with the controller IC, drives the primary side MOSFET. The feedback goes through an optoisolator to the control electronics. Some supplies do not have an optoisolator and instead use a control IC on the secondary side. These still can be modified, but it is trickier.

    The modification goes like this: after locating your feedback path, you must solder a potentiometer across one of the resistors. By adjusting the resistance of the potentiometer, you change the feedback range. If you insert a resistor in parallel, you can decrease the voltage; a resistor in series can increase it. Series resistors are more tricky because usually, a divider resistor must be removed or a track must be cut, and this runs the risk of damaging the supply.

    Another thing to be aware of is the fan usually runs off the 12V line, which can now vary from 7V to 9V. This can cause problems under heavier loads, because the supply may not be adequately cooled. You should take appropriate precautions to ensure the supply does not overheat.

    Enjoy.
  2. tom66

    tom66 Thread Starter Senior Member

    Joined:
    May 9, 2009
    Messages:
    2,613
    Location:
    Basingstoke, UK
    Some news. I've been able to disable the overvoltage protection circuit, which consisted of a small SCR set up in a crowbar configuration; designed to short the 5V rail in an overvoltage condition. By shorting its gate to the anode I was able to turn it off permanently. Then, I moved the potentiometer to the middle of one resistive divider. This allows me to adjust the 3.3V rail from 3.1V to 8.5V, the 5V rail from 4.8V to 11V, the 12V rail from 10.8V to 28V (massive!) and the -12V rail from -11.1V to -28.5V. I could probably get it higher, but the supply switches off at around 28V, not sure why.

    This is pretty powerful! The fan speed increases massively at higher voltages, because it is only rated for 12V. The output filter caps on the 12V rail are only 25V, so they may pop with 28V - similar for the 5V, 3.3V and -12V rails. The 5Vsb line stays at a constant 5.06V no matter what the outputs are, which is useful for powering a 5V fan without the fan speed varying due to voltage.

    It would seem this is a cheap way of making your own bench power supply which is adjustable over a very wide range. A cheap SMPS is available for less than £25. With some additional logic, it could be used to make a fully adjustable 3.1V to 28V power supply. Given that a zero to 30V SMPS power supply is around £200 here, this is a considerable saving. Sure, it's not as stable, probably has high ripple and poor load response, and it doesn't have a CCM mode, but these are usually minor issues for simple circuits. If desired, a linear front end, dropping only a few volts, could be used to create a zero to 25V power supply, giving it better regulation. The -12V rail could also be used as an adjustable rail, but it can't go to ground (only -11.1V), which might present problems for some circuits.
  3. thatoneguy

    thatoneguy AAC Fanatic!

    Joined:
    Feb 19, 2009
    Messages:
    6,356
    Location:
    Midwest USA
    Can you take a couple of photos to show what you've changed?

    I'd put this at "Medium" level, i.e. For those who have already built a working project of some complexity, like another power supply. Those who used an ATX power supply for +12V, +5V, etc could also do it, but there could be a big safety issue here if people who don't know a triac from a transformer attempt the mod.
  4. tom66

    tom66 Thread Starter Senior Member

    Joined:
    May 9, 2009
    Messages:
    2,613
    Location:
    Basingstoke, UK
    Last edited: Nov 12, 2010
  5. thatoneguy

    thatoneguy AAC Fanatic!

    Joined:
    Feb 19, 2009
    Messages:
    6,356
    Location:
    Midwest USA
    "I put 25 Volts to this card, and the chip didn't like it very much". :D

    Nice vid! Need to make a tripod to steady the camera a bit so it stays focused, but overall a good presentation and explanation.

    Thanks
  6. tom66

    tom66 Thread Starter Senior Member

    Joined:
    May 9, 2009
    Messages:
    2,613
    Location:
    Basingstoke, UK
    I think that chip was connected to the 5V rail so it got whacked with 11V. Needless to say it went up rather violently. First, three red hot holes were burnt through it, then it started burning and smoking... It melted some components on the other side, then one of the traces on the PCB melted so that was the end of that. :( I'm surprised the MOSFET drivers (which are probably Vds of 20V or less) survived given 12V was at around 25V.

    The camera I had was only a point 'n' shoot so no autofocus for videos, but good idea on the tripod.
  7. retched

    retched AAC Fanatic!

    Joined:
    Dec 5, 2009
    Messages:
    5,198
    Location:
    Baltimore, MD
    If you enjoy making videos, and doing tutorials like that, get yourself a camera.

    You can get HD videocameras for under $99US. Im sure they are around the same in your neck of the woods.

    And yes, get a tripod. They have the little 4 inch tripods available at some dollar stores.

    Not the best, but anything is better then motion sickness ;)

    Oh, and about the cutoff.. There is probably a thermistor that triggers a thermal cutoff. If you replace the fan, do you get more than the ~28v ?
  8. tom66

    tom66 Thread Starter Senior Member

    Joined:
    May 9, 2009
    Messages:
    2,613
    Location:
    Basingstoke, UK
    Yeah, like I have $99 US spare. Poor student here! :( I actually used a decent 720p 12 megapixel Pentax Optio P70 camera borrowed off a friend, but I compressed the video from 300MB to 45MB MPEG to make it sane to upload on a 2Mbps internet connection.

    You might be right about the thermal cutoff, but that was with virtually no load. It seems to be able to keep 25V out even with a relatively large load. I'm thinking that maybe the filter capacitor's dielectrics might be breaking down (as they are only rated for 25V) which could be shorting out the supply. I have some 35V 2200u caps here to replace the 25V 2200u's, so we'll see if we can get more volts out of it!

    The wires I have currently set up seem to short out the supply when the case is fitted (heh, bad soldering), so I'm going to look at adding some heatshrink on them to insulate them, then drilling a hole in the case to fit the potentiometer. Maybe I'll pick up a cheap £4 multimeter and stick it on the outside, wired to measure the output voltage, and a three-way switch to select the different 3.3V, 5V and 12V rails.
  9. mik3

    mik3 Senior Member

    Joined:
    Feb 4, 2008
    Messages:
    4,846
    Location:
    Cyprus, but now in UK (GMT+0)
    Have in mind that the power supply may become unstable because you change its feedback gain (loop gain).
  10. tom66

    tom66 Thread Starter Senior Member

    Joined:
    May 9, 2009
    Messages:
    2,613
    Location:
    Basingstoke, UK
    Did some tests with an oscilloscope and it's fully stable up to 28V. Additional ripple as you approach higher voltages. With an approximately 1A load it has <100mVp-p ripple, typically <50mVp-p.

    With the case on and no load, it will do 30.5V. The fan starts getting warm at this voltage - not good! So the cut-off point seems to be related to load current and temperature. I've found an ultra cheap 350W supply (which weighs less than the 120W supply - it's a piece of junk, tiny transformer, discrete diodes for the bridge, a 1000u cap on the 12V rail - absolute junk) which I should have some better luck with. It feeds the 5V straight in to a SMPS control IC, so it looks like I'll have to cut a trace to get it to work.
  11. mik3

    mik3 Senior Member

    Joined:
    Feb 4, 2008
    Messages:
    4,846
    Location:
    Cyprus, but now in UK (GMT+0)
    Test it for stability in a wide range of output current.
  12. tom66

    tom66 Thread Starter Senior Member

    Joined:
    May 9, 2009
    Messages:
    2,613
    Location:
    Basingstoke, UK
    I don't have any ability to do so.

    Your Mileage May Vary.

    At higher load currents, maximum output voltage drops. For example running a broken hard disk drive it is limited to ~24V. 30.5V is a peak output running only the fan.
  13. mik3

    mik3 Senior Member

    Joined:
    Feb 4, 2008
    Messages:
    4,846
    Location:
    Cyprus, but now in UK (GMT+0)
    If the voltage drops it means it is not well regulated or you have exceeded the maximum power.
  14. tom66

    tom66 Thread Starter Senior Member

    Joined:
    May 9, 2009
    Messages:
    2,613
    Location:
    Basingstoke, UK
    Probably maximum power. There are some current shunts and I suspect these limit the output voltage (by limiting the current.) Maximum power explains why putting the case on allows the voltage to go higher.
    Last edited: Nov 14, 2010
  15. Derrick_weedman

    Derrick_weedman New Member

    Joined:
    Nov 14, 2010
    Messages:
    6
    Location:
    Blufftion, Indiana
    Put together a bridge rectifier, not sure how to measure it on the multimert put it on ac first, the reading was 220v when pluged in a 110 outlet. Affraid I would fry the meter if I swiched it to dc till I found out.
  16. mik3

    mik3 Senior Member

    Joined:
    Feb 4, 2008
    Messages:
    4,846
    Location:
    Cyprus, but now in UK (GMT+0)
    You must stop playing with the mains voltage because it seems you do not know what are you doing. Voltages at this level are deathly.

    Also, you shall open a new thread if you want to ask anything.
  17. tom66

    tom66 Thread Starter Senior Member

    Joined:
    May 9, 2009
    Messages:
    2,613
    Location:
    Basingstoke, UK
    You could kill yourself!!

    Don't play with electronics above 30V unless you seriously know what you are doing. The project I'm doing is dangerous enough, but the power supplies' designers were nice enough to include discharge resistors across the filter caps, discharging them in 5 seconds after the mains is pulled out. You don't seem to have any such thing and run a serious risk of killing yourself.

    With a bridge rectifier you would measure DC. 220V is not unusual as the meter is probably not a true RMS meter. Remember RMS voltage is 1.4x lower than peak to peak voltage. Your meter is doubly compensating for a sine wave on top of the peak voltage it's measuring giving you 110×1.4×1.4 = 215.6V, which is about right.

    Be careful!
  18. themindflayer

    themindflayer Member

    Joined:
    Oct 29, 2010
    Messages:
    44
    Location:
    Pakistan
    excellent thread ....

    thanks
  19. tom66

    tom66 Thread Starter Senior Member

    Joined:
    May 9, 2009
    Messages:
    2,613
    Location:
    Basingstoke, UK
    No problem :)
Similar Threads
Forum Title Date
The Projects Forum Adjustable power supply too sensitive Oct 18, 2013
The Projects Forum -5V to 12V Four Terminal Power Supply Involving an Adjustable Regulator Dec 23, 2012
The Projects Forum Adjustable power supply Aug 25, 2012
The Projects Forum Adjustable power supply using variable digital pot Jul 4, 2012
The Projects Forum Slot Car power supply 12v 5-8amp adjustable Apr 17, 2009

Share This Page