Originally Posted by markus767
In the example case 1/6 might show what is audibly important but it doesn't show what is important
to set a EQ correctly.
Looking at higher resolution data really doesn't "hurt"
Here's a comparison how a EQ curve (blue) would look like for high resolution data and for lower resolution data:
Are you saying the spikes in the red circles are "important"? If you had an 8-band parametric EQ, would you use 2 of the bands to address those areas? If you did, they would be very narrow Q, low level filters. Do you think they would make an audible difference?
Obviously, if you have hundreds or thousands of filter taps, and you can correct those minor spikes, then only higher resolution measurements would provide the information necessary to set the filter taps. However, I'm not trying to write a program to use hundreds or thousands of filter taps. I'm just trying to set the subwoofer distance to eliminate a *huge* dip like the one at 48 Hz in your example. The granularity of my "tool", (the subwoofer Distance setting), is 1/10th of a foot or about 2", and the effect of those small incremental adjustments is 1 or 2 dB over a relatively wide band of frequencies around the crossover, (10 Hz above and below the crossover frequency.) 1/6 or 1/12 octave resolution is more than enough to show these variances.
OTOH, if one is trying to reverse-engineer Audyssey filter taps, then a higher resolution measurement would clearly be required. Good luck with that.