PWM is not all it's cracked up to be.
- Some cite lower power/heat. Negligible. Low speed fans consume a few dozen mA current
- More Stress on fan and bearings? True
- Can run closer to rail voltage- unimportant, if that little a drop was needed it was fairly pointless to have a control at all, just put a 2 cent diode in series if that's all you needed.
- Lower fan speed to get lowest noise? False, the fan will pulse at a higher speed than with some other methods because pulse width does not truely limit the non-linear current consumption from energization of the individual fan coils. The ultimate goal in quietest fan use is to keep the current surge from causing a sudden higher torque.
- Lower reliability? True, more parts to fail.
- Noisey. In some cases you get high pitched inductor whine.
- Larger and more costly.
PWM is one of those things that "seems" neat from an esthetic engineering sense but when it comes to true fitness for the purpose, it isn't.
For lowest noise use there is no substitute for current-limiting control. The key here is not to overthink what you're trying to do, which isn't to give the fan a lower voltage and expect it to then have a lower current when it's current is non-linear, it's to directly control the variable you want control over, which is what the motor itself is doing.
I should mention this is only in the context of lowest noise. PWM control (if your fan likes it) will do fine for medium fan speeds and general noise reduction. I just felt it worth emphasizing it is not really appropriate for a system that is already underclocked just to get even further towards a silent solution. If merely low noise was a goal there is no need to underclock at all.
