( Oh, heck, here he goes again with his off-the wall opinions )

Kirk Harnak typed:
> Perhaps even more important than considering whether one 
> can hear 50 kHz to 100 kHz material is asking "what does 
> removing such material do to the fundamental portions that 
> we do hear from 20 Hz to 20 kHz."

A very cogent comment, Mr. Harnak.  Maybe a similar situation 
in the RF domain is worth considering here. . . 

-----------RF speak begins------------

    We all know that FM sidebands go out "to infinity".  We also 
know that they are lower in amplitude the further (in frequency) 
you look above and below the fundamental of the modulated 
carrier, and that they are not useful once they drop into the 
noise, so at normal modulation indices we don't have to worry 
about high order sidebands.   Further, we know that band-
passing the modulated RF does, in fact, increase distortion 
of the demodulated audio. . . I don't know what the exact 
numbers are, but I think I read somewhere that removal of 
anything closer in than the 3rd order FM sidebands increases 
distortion of the recovered audio to the point where it is audible 
to the untrained human ear.  
    On the other hand, we all know that _not_ bandpassing the 
RF creates a complete inability to demodulate an FM carrier 
in the real world.  That's why we have receivers that 'tune' 
to a certain, small chunk of the RF spectrum, and filter out 
the other stuff. . . a bandpass filter.  The wider this filter, 
the more prone the system is to noise and interference 
contained in the demodulated audio, the narrower this 
filter, the more distortion of the desired signal we endure.  

    So, there's a 'sweet spot' with respect to filtering out-of-band 
stuff, and if you get more aggressive than that, "bad things 
happen." 

    Now, I'm not trying to say that audio acts exactly the same 
way, but I think there are similar 'kinds of things' going on in 
the two domains.  

------------RF speak ends--------------

    Similar things happen when we bandpass complex audio 
waveforms.  We remove out-of-band noises which do not 
contribute to the listening experience, but at the same time 
we change the in-band energy in ways that are not 
always subtle.  An example: removal of harmonics at 
frequencies that our ears cannot resolve in a test using 
sine waves, but  _may_ be able to sense as harmonics 
of inband information because they are 'synchronous' 
with that in-band information, giving the ear clues to 
assist in detection.  I believe that you can sense a change 
in the 'edginess' of the sound of an instrument using 
frequencies that are much higher than you can sense 
as single tones.  I also feel that sine wave testing up near 
20 kHz in earth's atmosphere is a real iffy idea at best, 
because short wavelength reflections have a chance to 
create comb-filter effects that greatly affect steady state tones, 
but don't affect complex audio waveforms in the same way.  
There may even be propagation linearity issues caused 
by the way the gases in our atmosphere act when 
subjected to changes in pressure that are that rapid.  
That's beyond my area of expertise.  

    My approach has always been to bandpass the audio 
as early in a system as possible, to minimize exercising 
non-linearities in subsequent system components with the 
out-of-band information.  Supersonic junk and subsonic junk 
are just that. . . junk.  We don't need to deal with record warp 
and turntable rumble anymore, but there's still subsonic information 
from many other sources, especially with live microphones.  
The question is "WHERE?"  In a non-broadcast environment, 
I like simple 1-pole filters at 25 Hz and 30 kHz.  Easy to 
accomplish, and in-band phase shift is minimal.  Narrower for 
broadcast purposes, and _much narrower_ in microphone 
circuits for speech, to reduce sibilance effects and possible 
heterodyning with supersonic oscillations in unstable opamps 
in downstream equipment and subsonic excitation of possible 
downstream capacitor non-linearities.  

     After I die, folks who disagree will have to pull a lot of 
polypropylene capacitors out of my systems   (c;  .


Grady Moates, April 2003 

Owner, LOUD & Clean Broadcast Science
  Voice        800 946-7007 
  Facsimile  800 529-5648 
  web          <www.loudandclean.com> 
  email         grady.moates@loudandclean.com