On Wed, 17 Aug 2005 05:38:32 -0500, Analabha Roy
<daneelNOSPAM@physics.utexas.edu> wrote:


>> Set up capacitive / resistive loads on each rail,
>
>
> How? do I get a breadboard, put some resistors and capacitors in them, and
>use that? How will I be able to connect that with the power port of the
>PSU? Do i have to unscrew and remove the chassis or something?

It'd have to be quite a hefty "breadboard", considering that
the load would require 400W heat dissipation... put in
perspective, you would need several very large heatsinked
resistors, a heat transfer plate, and fairly high velocity
fans, or an oil bath... whatever your chosen method of
dispersing the heat. As for the capacitance you'll need,
rought guess would be a 10,000 mfd per rail (varies per
system, that might be a rather conservative figure for a
system actually using anywhere near 400W which in itself is
hard to do with a PC, but more importantly some systems may
actually use the majority of the 12V rail.

How to connect to the power port? Buy the mating molex
connectors. Setting up a testbed isnt' the kind of thing
one can usually do with bits in their spare parts bin,
unless they have a pretty generous stock of parts. You do
not need to open the power supply at all, unless you also
wanted to measure temps.

Frankly all of this is a waste of time, as opening the PSU
probably voids the warranty then you can't return it.
Testing until it fails is also a bit of a loss, really only
useful to quality fhe design for larger scale deployment
over many systems.


>> hook up a
>> scope and a few multimeters, vary the load rapidly like
>> modern CPUs/GPUs/etc do, and run some tests for a few
>> months, recording voltages, recovery time, ripple, etc.
>> Then test shutdown thresholds like overcurrent, overvoltage,
>> overheat, fan lifespan and response... There is no way you
>> can do a simple "it's good long-term" type of qualification
>> for it in a day or two.
>
>But isn't it possible to just verify the maximum power threshhold of the PSU
>without too much testing?

No, you could get a peak sustainable value that's useful to
the extent that you know if the rails would drop with that
degree of system load, but it won't tell you if there's
potential for damage from massive ripple, won't tell you
what the PSU does in a fault scenario, and doesn't tell you
how long it will run at that output level. Even a poor PSU
may put out 300W for a day, maybe even weeks to months.
That's the worst scenario, because then after the warranty
is up it may fail, or even if warranty remains you generally
aren't covered for any OTHER parts that could be damaged.


Look at it another way- if there was no significant drawback
then how would the other manufacturers stay in business
selling them for far more when they all (until recently)
just looked like boring grey metal boxes?