FAN A: +70°C ∆T at 10 dB
FAN B: +50°C ∆T at 30 dB
Your ranking system gives both of these fans an identical NTT rating of 9.0. But anyone with sense would choose FAN B over FAN A. So what good is your rating? It only works when you hold one value (either noise or ∆T) constant between both fans; but if you do that, there is no need to do any additional arithmetic, because you have a single frame of reference. You might as well simply quote the other variable (the one that varies between the two fans for that reference level):
FAN C: +41°C ∆T at 25 dB
FAN D: +48°C ∆T at 25 dB
SPCR deals with this in a very good way; they create three reference sound levels (e.g. 11 dB, 16 dB, 21 dB) and show the ∆T for each. Another way would be to create a more detailed noise-to-temp curve for each fan as it varies along its RPMs, and graph them all together (this is essentially just a richer visual representation of SPCR's coarser review method). You cannot reduce the matter of interest to a single value, however, since it's an inverse relationship with a non-linear curve.
If it WAS linear, the relationship would be:
∆T = R x (dB) + i
You can immediately see that even an inverse linear relationship is problematic since both the slope R and the non-trivial intercept i are germane to our concerns. On a dB vs CFM graph the intercept is zero, so we can simply refer to the noise/CFM ratio as a good comparative ratio between fan designs. Of course, in reality, even noise vs CFM is not actually linear, so that is better represented graphically.
In any case, directly adding noise and temps is physically meaningless and does not reflect any kind of real-world relationship between these values. More critically, it is simply not useful.
Edited by Gabedamien - 8/31/12 at 3:32pm