Connect with us

Rating of PC power supplies?

Discussion in 'Electronic Design' started by Terry Pinnell, May 29, 2005.

Scroll to continue with content
  1. I'm not familiar with switched mode power supplies so hope this
    doesn't prove to be a dumb question. I dusted off an old '200 W' ex-PC
    power supply yesterday and was playing with it with a view to using it
    for some motor tests. (For the time being I settled on a set of 4
    NiCads instead, but I may pursue it later.) Its spec says that its
    '5V' output is rated at 20A. Arbitrary tests gave me these results:

    Current Voltage at load terminals
    ------- -------------------------
    5.1 A 4.77
    7.5 A 4.51
    11.0 A 4.20
    13.0 A 4.02

    I didn't go further, but even at 13A isn't that voltage rather low? I
    wonder what it would drop to at close to 20A? What is regarded as an
    'acceptable' voltage from the '5V' output of these types of supply
  2. One day Terry Pinnell got dressed and committed to text
    G'Day Terry
    Pity you didnt check at the output on the PSU, could have shown up any volt
    drop in your load leads. The regulation feedback comes from those
    Then again perhaps I'm wrong (as usual :)
  3. Franc Zabkar

    Franc Zabkar Guest

    IME most PC PSUs regulate on the basis of a weighted average of the
    +5V and +12V rails. If your +12V rail is unloaded, then the +5V rail
    will also not regulate properly. I'd check the capacitors on the +5V
    rail (or add an external cap for testing purposes), and I'd measure
    the +12V rail in all four cases.

    - Franc Zabkar
  4. Actually I did, Rheilly - just didn't include the correct column in
    the posted results! Here's the fuller table, which as you rightly say
    gives the proper picture:

    Current At PSU output At load terminals
    ------- ----------------- ------------------
    5.1 A 4.96 4.77
    7.5 A 4.78 4.51
    11.0 A 4.60 4.20
    13.0 A 4.50 4.02

    My 'PSU output' was at the end of one of the red '5V' output wires,
    about 30 cm long, the first location accessible without opening the

    Wiring from there to my load (and back to the PSU 0V cable, a similar
    distance from case) therefore totaled about 0.04 ohms.

    So, much better, with 4.5V at 13A. Not sure how to estimate 20A. Do
    those numbers look acceptable?

    BTW, correct operation only started with a load of something over 1A
    or so. Presumably a characteristic of this type of PS?
  5. Yes, that's correct. You can Google around for information on how to
    successfully use PC PSUs for general purposes.

    Also be aware that there is more than one 5V output, at least if we're
    talking about a standard ATX supply, e.g., I don't know whether your
    unit was rated for 20A at *each* 5V output, or for them all put together.
    They should all be coregulated, indeed they might even all just be in
    parallel; but it sounds like the issue you're facing has something to do
    with IR drop between the sense point and the load, so it makes a difference.
  6. Thanks. Yes, there are 3 '5V' outputs, apparently in parallel. I'll
    check whether connecting them makes any significant difference.
  7. Guy Macon

    Guy Macon Guest

    No-name PC power supplies are notorious for being junk. Buy a
    lab supply, build your own supply, or buy a PC PS from PC Power
    and Cooling (they make very good PC power supplies).
  8. Rich Grise

    Rich Grise Guest

    I'd parallel all of the black wires as well; forgive me if I'm being
    redundant here. :)

  9. No harm in trying it, so I will! But without any expectation that it
    will affect any measurements. I'm pretty sure I recall when this PSU
    was inside my old PC that several of the output 'sets' were floating
    loose in their plugs, presumably awaiting service for additional hw.
  10. Franc Zabkar

    Franc Zabkar Guest

    You *must* load the +12V rail with a dummy load, otherwise you will
    *never* achieve proper regulation on the +5V rail.

    See this ATX example for an explanation of how a typical PSU

    The regulation formula is:

    5.0V * 3K01/(3K01 + 3K01) = V12 * (3K09||150K||5K6) / (27K +
    + V5 * (3K09||150K||27K) / (5K6 +

    This reduces to:

    2.50 = V12 * 0.0678 + V5 * 0.327

    For this PSU, if the +5V rail increases by 50mV, say, then the +12V
    rail will fall by 241mV.

    If you want to guarantee a stable +5V output, then remove the +12V
    sense resistor and recalculate the remaining +5V sense resistors such
    that the midpoint of the potential divider is +2.5V. I have modified
    several PSUs this way. I have one which I have reprogrammed for +6V,
    and another for +13.8V.

    - Franc Zabkar
  11. legg

    legg Guest

    Only one of the output and return terminals will be sensed. Use all
    releventterminals in parallel to remove harness drop from your

    You can identify the sensed terminals by visual inspection. These will
    have two wires crimped into a single power pin.

  12. Pooh Bear

    Pooh Bear Guest

    Usually spare drive power connectors.

  13. Pooh Bear

    Pooh Bear Guest

    I'd expect it to range from 5.1 to 4.9 volts approx. They're normally not
    too bad. Shouldn't really be more than 200mV out of spec.

    Bear in mind that it's a multiple output supply and an absence of load on
    the 12V output is likely screwing things up.

    As another poster noted, these supplies *have* to sense a combination of
    both the 5V and 12 V outputs so no load on the 12V is bound to be an
    issue. The 12V is probably high ( due to no load ) and therefore dropping
    the 5V low.

    The -5V and -12V outputs are low current and just rely on the reflected
    voltage. They won't need loading.

  14. Pooh Bear

    Pooh Bear Guest

    That works fine if you have the schematic to calculate the values. ;-)

    I suspect Terry's old PSU is likely an AT type btw.

  15. Many thanks, and for the other follow-ups. I'll load the 12V output
    and re-test.

    Note that my later post prompted by Rheilly gave the more relevant

    Current V (At PSU output)
    ------- -----------------
    5.1 A 4.96
    7.5 A 4.78
    11.0 A 4.60
    13.0 A 4.50

    I'd be more than happy with even 4.8V at 15-20A. Crude motor tests
    apart, that would be handy if I ever build any more stuff from my
    ancient stocks of TTL. (Don Lancaster's 'TTL Cookbook' recommends +/-

    Must say I'm impressed at the performance of these things. I expect
    there are downsides in hf noise, but they're so light and compact in
    comparison to a conventional transformer-based equivalent. I made a
    heavy duty 20A car battery charger years ago, still in use. OK, it's
    12 not 5V, but I have to wheel that monster out on an ex-shopping
    trolley <g>.
  16. Franc Zabkar

    Franc Zabkar Guest

    Not necessarily. The most common PWM controller IC is the TL494 or its
    equivalent, KA2500B/C. All you need to do is to confirm the voltages
    at the inputs to the IC's error amp. One input will usually be 2.5V (=
    internal 5.0V reference divided by 2), the other input will be derived
    from the +12V rail and/or the +5V rail via a resistive potential
    divider. All you need to do is to recompute the resistor values in the
    feedback network so that the target voltage, eg 13.8V, produces 2.5V
    at the input to the error amp.
    The two that I modified were both AT types. I had the schematics for
    neither. Only very minor reverse engineering was required.

    - Franc Zabkar
  17. Nice post, thanks. As you see from my other reply. I'm off to the
    shed/workshop to try it soon! Will probably take the easier approach
    of trial and error loading, until I get a satisfactory minimum, and
    will report back on results.

    When I do get around to opening the case of such a unit, that
    schematic and notes will be invaluable.
  18. SioL

    SioL Guest

    I have experimented with these a bit, albeit an older variety (AT type) of todays
    power supplies. What I found is that 5V is usually pretty accurate, while
    12V is somewhat flimsy. This particular PS was only sensing the 5V
    output, relying the others would be more/less close to the design ratio;
    it can be off quite a bit for +12V and varies a lot with load.

    The only output that had to be loaded was the sensing output, that was 5V.
    My reason to play with these was conversion into 10-15V variable voltage PSU,
    which involved cutting sensing track from 5V and running it to 12V output
    via adjustable voltage divider. Loading was now only required for 12V output
    and fan did a good job of that (required some LC filtering to make it quiet electrically).

    Made for a nice encased dirt-cheap hobby power supply with current protection. Some
    of these PSU's had chromed enclosured, looked fairly good.

    Hopefully someone will benefit from this.

    Take care with high voltage, can be painfull/lethal!!
  19. Thanks, very useful.

    I've repeated my tests after connecting all but one of the 5V (red)
    outputs, and all but one of the 0V (black) outputs. This time I was
    also more careful where I measured my voltage; I used the unloaded
    5V/0V wires. Happy to say that I now get pretty solid 5.1 V right up
    the range. And, like you, I did not have to load the 12V output, only
    the 5V.
  20. As per my reply to SioL, I did not have to load the 12V output, only
    the 5V.
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day