A few days ago I posted about a PWM controller in a WLAN router that emitted acoustic noise and got a few good hints. Unlike I wrote at first, however, the effect is not a thermal one but has to do with the fact that the power consumption of the device changes after boot-up of the embedded processor. I checked the switch-side node of the coil with an oscilloscope, and indeed when the things makes its noise the PWM pulses are all over the place. No two traces look alike. To see if this had anything to do with the output capacitance I soldered various caps (electrolytic and foil) in parallel to the output but found no change at all.
Now I'm going to have to look at their feedback loop and see if I can do anything about that.
robert