Whistling cooling fan on PM 8500?

[noidentity]

My PM 8500 has been driving me nuts for many years now. Occasionally when the room temperature/CPU load is just right, the fan will start to whistle at around 532 Hz (slightly above a C5 note).

Often it occurs when I'm holding the mouse button selecting something, and then goes away when I release, though delayed as the fan takes a few seconds to change speed every so slightly. Or the other way around, making the sound, but goes away when I hold the mouse or do something that uses the CPU more. If I cover up the back vent with my hand when it's making the sound, it goes away, so it's very likely related to fan speed. I'm thinking that the power supply's voltage output varies slightly based on CPU load, which affects the fan speed. I've removed the fan several times from the power supply and it definitely doesn't have any visible thermistor, so it's not varying speed based on temperature. I've noticed that they have the fan almost right next to the four VRAM SIMMs, so they might be interacting with the air flow of the blades as they spin past.

I've got a Sonnet 400 MHz G3 CPU card, the one that uses a copper processor and runs really cool, and only one hard disk active, so I'm not putting much of a load on the power supply. I run OS 9.2.2. When I've had the power supply pulled apart, I've never seen any bulged capacitors or leakage either, and I have kept the inside nearly dust-free with regular cleaning. Yesterday I even spun down the hard disk while it was whistling, and it continued. Room temperature is around 79 F, and I believe it doesn't do so when it's several degrees cooler.

Anyone dealt with this kind of thing before? I was thinking of researching how air resonates, to see whether I could find any structures that might cause this. Thanks.

 

[noidentity]

I've made progress on this. I found that the Delta power supplies used in the PM 8500 do in fact have a fan speed controller. The fan's + is connected to +12V, but its ground doesn't go to ground. I also found the thermistor nearby (lower resistance = higher temperature). I connected two wires to it and brought them outside the case. With a 100K variable resistor across it, I am able to speed up the fan and make it hit that resonance frequency. Interestingly, it can be made to run a lot faster than normal (I found the same when I plugged the fan into straight 12V; it moves lots of air). Apple made such a big deal about the variable-speed cooling fans in much later PowerMacs, I didn't expect such a thing to be in a "lowly" PM8500. For some reason I find this really surprising.

I'm thinking my simple solution for now will be variable resistor with a toggle switch across the thermistor, with the resistor set to slightly speed up the fan. Then when the fan starts to make that awful whine, I'll just toggle the switch, which will either speed it up slightly or slow it down, depending on whether it was closed beforehand. Either way it will get away from the resonance speed and allow me to stay sane.