Converted atx psu whistle

[sureshot]

Wander if anyone can throw some light on a converted atx PSU issue.
OK so I followed the normal route and loaded the 5 volts rail with 10 ohms 10 watts resistor to use the 12 volts output, only there is an audible whistle, whilst it does not affect my use of the unit it is annoying, I have found an additional load of the same resistor as above on the 3 volts rail cleans up the noise.

Can anyone explain this behavior, I have googled this to no avail, and checked for bad capacitors and can not find any, if i leave the load off the 3 volts rail will it do any damage other than the annoying whistle, thanks if anyone can help.

 

[KrisBlueNZ]

The likely reason for the whistle is that the supply is running with such a small load that even at minimum duty cycle, the controller is skipping cycles.

Under normal load, the controller drives the switch and main transformer at a fixed frequency, for example 80 kHz, and varies the duty cycle of the drive signal to adjust the amount of energy delivered into the transformer to match the load power and maintain regulation.

The controller has a certain minimum duty cycle, for practical reasons. If the load power is less than the amount of power delivered at the minimum duty cycle at the nominal switching frequency, the controller needs to reduce the energy delivered to the transformer to maintain regulation, but it can't reduce the duty cycle any further, so it starts to skip cycles.

Once this happens, the drive signal starts to include components at submultiples of the switching frequency. In an extreme case with very low load, the controller might skip seven of every eight cycles, and the switching frequency would drop to 10 kHz, well within the audible range. But even if the load is just slightly too low, audible frequency components will be produced - if the controller is skipping one of every eight cycles, that also generates frequency components at 10 kHz and will be audible.

So there are several ways to avoid the problem. First, you can ensure the minimum load never drops below the threshold where the controller starts skipping cycles, by adding load. It doesn't matter which rail you add load on. Increasing the load on the 5V rail will work as well as adding load on the 3V rail.

It might be possible to change the controller's minimum duty cycle or some other parameter(s). I can try to look into this if you tell us the markings on the controller IC.

You might be able to reduce the noise by spraying aerosol lacquer into the windings of the transformer and any inductors you can find, to reduce vibration, or put the power supply in a box (as long as that doesn't cause it to overheat).

The cycle skipping isn't a big problem and it won't do any harm to the power supply. It may correspond to a higher than normal ripple amplitude on the outputs though.

 

[sureshot]

Thanks KrisBlueNZ for the very in depth explanation, I had an incling it might have been load but was just not sure.

Its is in a cupboard at the moment so I don't notice the noise to much.This is a very cheaply made old atx, but thought a of some use rather than the bin, so first I will try and increase the minimum idle load see how that goes, I'm using this on an old cb radio and the TX and RX audio seem clear of any noise, although I don't have a scope measure this. Thanks again much appreciated, if I get to the open it up stage again I will make a note of the controller IC and snap a picture or two.
Cheers.