I really like your sense of humor.
I don't care about the fan noise itself. I experimented with 140mm case fans and let them spin at 200 rpm. Absolutely noiseless. It's about the micro-housing resonance that such a fan always causes. Sensitive electronic components are given a massage. The PWM control ensures that the current is regularly switched on and off in the high-frequency range in order to maintain the rotation. Last but not least, a fan causes dust to be blown up. If you make it long enough you can make a winter blanket out of it.
Thanks for this and the tip. I think about that.
Fair enough about the fan vibration, thought there is so much rubbish from the processing the fan vibration wouldn't be noticed.
I can't recall if the mobo fan PWM controllers can be disabled, if left to full, they will be active at all times making noise.
There is much to optimise in the server, the most difficult are the mobos and the power supplies. On the latter, high frequency switching in the MHz takes it out of the audio band, but creates more harmonics and higher freq conducted noise in the 30MHz region. You would have to switch higher than this to stop conducted noise, but designing power transformers at this speed is certainly a challenge....
On LPS, the use of tight regulation supplies may not be needed for devices like the EtherRegen since it will accept 7-12V, or the JCAT cards 4.x to 5.1V are acceptable, so they can be simple transformer, diode, cap combos without regulator contributing noise.
For a single 12V rail mobos, I need to check the tolerance on the mini ATX mobos, since on board regulation takes care of the rest, so a simple LPS is all that is needed.