Why the howl ?

Basic theory of PWM control:
http://www.pcsilent.de/showInfoPage.do?name=funktionpwm

As you can see the voltage to the fan is either 0v or 12v, nothing in-between. Every time 12v is applied the fan makes a little kick which is audible to some degree dependent on the fan construction and how it is mounted. When lots of these voltage changes are applied in sequence the sound from the kicks merge into a tone with a frequency equal to the repetition rate of the voltage change aka pulse repetition frequency or prf. For fan control the prf is usually between 10Hz to 400Hz this tone is clearly audible as the howl. The easy fix for this would appear to be increasing the prf to a point beyond normal human hearing, somewhere above 20,000Hz, the problem here is such high frequencies cause heating in the fan drive electronics somewhat negating the efficiency gains pwm brings and probably shortening the useful life of the fan.

There are a couple of tricks that can be used to reduce the pwm noise:

1) Change the shape of the 12v pulse from a vertical line to a more gentile slope or curve:
http://ww1.microchip.com/downloads/e...tes/00771b.pdf

... nice idea but for best effect it needs to be optimised on a per fan basis and in my experience can actually make the fan noise worse.

2) Continually change the prf so the noise from the kicks no longer makes a constant tone. Good idea in theory, I know someone is looking into this but I've yet to try it myself.

In an ideal world I'd know enough about electronics to leverage the ideas behind switch mode power supplies to produce a variable buck or boost converter. This would be efficient, able to provide high currents and have zero pwm noise. It's on my list but as the fan header on my mb is proving quite sufficient at the moment it's not a high priority.