Home General Asylum

General audio topics that don't fit into specific categories.

Your comments on the Shannon theorem

Christine, the ear does a mechanical filterbank that does an (odd sort of) time frequency analysis.

It is not very sensitive at all above 20kHz in children, and not sensitive at all in older people.

Filtering a signal to a 20kHz bandwidth is, therefore, not much of an issue, for several reasons.

1) If you don't close-mike something (see violin discussions elsewhere) you don't see very much energy up there, and that energy is very often below absolute threshold
2) Air propagation, unless it's VERY dry, tends to suck up high frequencies rapidly.
3) Quantization error is nothing more than a noise floor. Your two-tone signal suggests that you think digital is not linear. It is, unless somebody did it wrong. BUT when you say "vary by time and amplitude" you're modulating the signal, and therefore spreading its bandwidth, perhaps lots depending on what kind of "vary"iation you're doing. What's more, well, no, you will NOT see the kind of results you expect in terms of dynamic range, digital acts just like bandlimited analog. The noise floor is a bit more controllable in digital, that's hardly a deficit.

The only thing you'll see in a digital signal is bandlimiting. You will see that in an analog signal as well, but depending on the device, perhaps less so. (Now, put a square wave through an LP, go ahead, you'll see it's often even worse.)

Square waves, etc, have frequency components that go to infinity, but which roll off rapidly, square waves by 1/n relative to harmonic, triangles by 1/n^2, and so on.

And you need to show, using a good, linear transducer and system, that the harmonics over 20kHz are audible. To date, only one questionable experiment that wasn't dismissed due to transducer nonlinearity has suggested as much.


This post is made possible by the generous support of people like you and our sponsors:
  Sonic Craft  


Follow Ups Full Thread
Follow Ups


You can not post to an archived thread.