[[email protected]]

It is based on analyzing a lot of natural recordings with Cool Edit and a
Pentium 133 computer to see if a "natural sound characteristic" existed. The
idea was an expansion of something a dane at an AES meeting in Radiohuset
about restoration explained about comparing a new recording of a string
quartet with an old to get data that would be useful in recovering the
original frequency content, I'm very sorry that I can not remember his name.


this is an interesting topic.

I was under the impression that the average spectral density of most music follows the 75us FM radio de-emphasis curve. It is flat from 20 Hz to about 2.2 kHz and rolls off at 6 dB per octave above 2.2 kHz. And I do find that recordings that violate this on the high side do seem to sound harsh.


Then another tib-bit.....

It all depends on the type of analyzer you use.
A so called real time octave based audio analyzer has "bins" that get wider as you go up in frequency.
A "normal" spectrum anlyzer has "bins" that are of a constant bandwidth.
This makes no difference when you are analyzing tones.
But when you are analyzing distributed wideband spectrums, it can make a difference.
For example, pink noise appears "flat" on an RTA and slopes down on an SA. White noise appears flat on an SA and slopes upwards on an RTA.

Having fun yet?

Mark