If you ask me, the noise is not necessarily electronics related!
It could just be it's fan being whiny.
FYI, computer PSU's regulate the fan speed depending on the load (they increase/decrease it's voltage to do so) so that would explain why it's more noticeable at higher loads. You could try replacing it.
Could be fan PWM or blade or bearing noise, but perhaps unlikely. (Usually sounds different, or others more likely. Not ruling it out though.)
Acoustic emissions = something vibrating. A fan is the obvious source. But inductive components can vibrate, due to the varying magnetic forces on the windings and the cores.
Telcommunications power supplies, which have to meet acoustic emissions specs., commonly glue transformer core halves, to stop any movement. A near-invisible speck of dust, between two unglued core halves, acts like a pivot. When energized, the core halves can rock back and forth on the speck, causing an annoying whine.
For really quiet operation, the windings are also glued, to stop any movement there.
Not widely known, but well understood: magnetic materials are usually magnetostrictive, i.e., they shrink (by some ppm) when a magnetic field is applied. This is magnitude dependent, thus the acoustic signal is twice the magnetization frequency. Or, a bias must be present to make a minor oscillation apparent.
The latter is the most common case: the control loop becomes unstable, due to increasing capacitor ESR (and decreasing value), causing oscillation.
The oscillation is audible through a few possibilities:
- Power transformer (ferrite cored)
- Filter choke (ATX supplies usually use #26 or #52 powdered iron, which is lower magnetostriction than the ferrite core transformer, but it's still nonzero)
- Film capacitors (very unlikely in electrolytics, but quite common in film capacitors, which exhibit the analogous electrostriction property; a film capacitor is typically used for coupling and/or bypass in the half bridge topology)
if all fails , then change the output caps with i slight bigger value. that will change the frequency ,hopefully to a non acoustic one.
explain how?
Old capacitors have high ESR, the output capacitors being the most critical. ATX supplies are usually voltage mode, so the capacitor ESR is a necessary zero in the loop stability criterion. Too much and, not only is the output ripple way beyond spec, but the loop goes unstable and you usually get a whine at the loop cutoff frequency, which is usually pretty low, a few kHz. So, there you go.
Tim