In article <434b90af$0$10578$3b214f66@usenet.univie.ac.at>, Martin Paul <map@par.univie.ac.at> wrote:
>All of my Blade 1500 workstations (blue) generate a strange kind of howling
>noise. Taking a close look with the case opened, I know now that it's
>caused by the case fans, which accelerate/decelerate about once or twice
>per second - one can hear that, and even see it when looking at the
>fan blades.
>

I had a 2500 (same vintage as the 1500)
that used to make a sorta clicking noise.
I think it was the PWM frequency aliasing with the
fan's rotation frequency.

I soldered a small electrolytic cap (note polarity, or it
turns into a firecracker) to the fans leads where
they entered the connector (unlatched barb,
pulled out pins, soldered, and replaced).
It smoothed out the voltage, and all the noise went away.
If you could find the right connectors you could
even put it inline, and plug it in (and remove for a
warrantee return, if needed (or just snip off
the cap, if you do it may way))

The cap I used was about a centimeter in diameter.
The value isn't critical, it just needs to be "big enough"
to smooth it out.
Make sure it's rated for more than 12 volts.


>One can see it in prtdiag -v output, too. I used:
>
> while test 1 = 1
> do
> date
> prtdiag -v | egrep 'rpm|Ambient'
> done
>
>The temperature is constant, but the fan rpm of both case fans is changing
>frequently. I guess the problem could easily be solved in software, by
>modifying the fan speed control. Is it possible to do this via any Solaris
>command (picl/*fru*/whatever) ? I'd really like to check whether e.g.
>setting the fans to a fixed rpm would resolve the noise problem.
>
>If nothing else, I'd like to know if anybody else has this problem.
>
>Running the fans on full-speed all the time would probably be less
>disturbing than this pulsating sound.
>
>I'm up-to-date on patches, and I just updated the OBP to the most recent
>version, too (I hoped this could help, esp. after seeing "4816966: Need to
>change Fan PWM Freq in OBP for Taco" in the OBP patch README, but to no
>avail).
>
>mp.