On Thu, 28 Jun 2007 09:32:10 +0200 (CEST), Christian Emmerich wrote: > >Thus this algorithm actually return sane RPM values for you? Now I > >remember why I did not use this - Here the "Count" rises linear, which > >makes "RPM =(1.5?10^6)/ Count" *lower* when fan speed increases. OTOH it > >might be the fan is just too small and thus rotates too fast to be > >reliably measured.. > > During implementation i tested different values for pwm here are my results: > pwm1 fan1 fanspeed ( using: fanspeed =(1.5?10^6)/ fan1-value ) > value value > 13 2600 ~570 rpm > 14 1900 : > 15 1900 : > 23 1100 : > 34 770 ~1950 rpm > : : : > 90 385 ~3900 rpm > 95 375 ~4000 rpm > 100 365 ~4180 rpm > 105 360 ~4230 rpm > 110 345 ~4360 rpm > 115 345 ~4360 rpm > 120 335 ~4500 rpm > 125 330 ~4630 rpm > > > So if i start my pc the pwm is at ~32 and the fan is turning at ~1800rpm. If i increase > the pwm value the value of the counter decreased! Yes, this is expected. The counter value is proportional to the time the fan needed to achieve one revolution, so a greater counter means a slower fan and vice versa. > So if the pwm value is ~110 the van turns at >4000rpm > and the fan1 value is ~365. > The pc i use is a barebone by AOpen with an special mainboard and fan, both comes with the barebone. It seems to work OK. So the conversion from count to RPM should be moved to the kernel driver. Thanks, -- Jean Delvare
