Standard deviation in speaker measurement -

[TLEDDY]

On the preference ratings, what is the standard deviation in those measurements?

From my background in Laboratory Technology, a blood sugar reported as 100 mg/dl with a procedural SD of +\- 3 mg/dl could have an actual value of between 97 and 103 that is perfectly acceptable precision.

Just curious… I would hazard that the SD in electronic measurements are much more precise than biochemical reaction measurements. Sort of like measuring the length of a one inch wooden dowel with a yardstick or a precision micrometer. When I started working in a hospital laboratory in the 1950s, one method used a visual color comparison with a calibrated (well, sort of) glass standard with the color chemical reaction of the patient sample in a glass tube. For sugar analysis with that method, an SD of 10 or 15 was pretty good. Imagine the variables in a human differentiating color, much less the entirely manual pipetting of blood and reagents!

 

[africanus]

In statistics, samples have a property called statistical power (https://en.wikipedia.org/wiki/Power_of_a_test) that depends on the nature of the test, sample size and the assumed (a priori) shape or distribution function of the variable under investigation, among others. Since audio measurements are performed just once and only on one unit of equipment, there is no information enough to calculate things like standard deviation or significance level. Preference ratings are the result of applying a predefined formula to those one-time measurements so their sample size and degrees of freedom are the same of those of the measurements on which they are based.

 

[Frgirard]

It would be interesting to measure the passive speakers in pairs to see the deviations.
In the 90s, at Cabasse, no passive loudspeaker sounded the same as the other (unlike the VT or VTA range) because of the disparity in drivers.

AND

It would be interesting to measure the active speakers in pairs to see if the calibration is real (+/- 1dB) or advertising.
Implicitly it is recognized that the drivers do not have the same sound.
For a brand like Neumann the calibration is real (Neumann don't equalize the mid)

Adam did not equalize his drivers on the former S5XV.

 

[TLEDDY]

Thank you!