Zach points out that "16 bit 44.1khz audio is limited". Of course it is, and the limits are easy to demonstrate. Nothing can be recorded above half the sample rate (actually a little less than half the sample rate...a little beyond 20khz). And in theory nothing can be recorded below the 16 bit limit of 96db (6db of dynamic range per bits 16(bits)x6(db)=96).
Actually there's a way around that last one. "Dither" random noise at just above the -96db level, which keeps the dac from ever "switching off" (as it would with audio below -96db otherwise) allows audio to be recorded below the "theoretical limit" of -96db. Another technique called "noise shaping" takes sampling noise (generated in any digital format) and moves it from the midrange where the ear is most sensitive, above the audio range (above 20khz). Taken together, these techniques allow what is ostensibly a 16 bit format to record resolution behond the "16th bit".
I know...another long winded lecture from Walker. My point is that yes, 16 bit 44.1khz is "limited". But so is human hearing. 16 bit 44.1khz was chosen because it's a very close match to the limitations of our hearing. -96db is not only quieter than the quietest of our listening rooms, it's quieter than any real recording studio. 20khz response is beyond the hearing range of almost everyone. And even if you can hear a discreet tone at 20khz, the absence of frequencies above this point means little, because music isn't a "discreet tone". It's a complex waveform composed of many, harmonically related frequencies. Real musical "notes" (the fundamental tones) occur almost entirely below 2khz. What's above 2khz is harmonically related "overtones". Well the ear is most sensitive in the 500-2khz range, because that's where voices "live". Surprise...God tuned our hearing to be most sensitive to the sound of our own voices. Makes sense!
Well our ears aren't laboratory instruments. Unlike test gear, they are quite incapable of hearing all components of a complex waveform simultaneously. Play a tone at 1khz. Now play a tone at 800khz that's 15db quieter. Now play the 1khz tone at full volume again. Then play both tones together. IT WILL SOUND THE SAME WHETHER THE 800hz tone is present or not, because your ear can't hear a soft tone of lower volume in the presence of a loud one of similar frequency. This is why analog noise reduction systems like dolby seem to provide constant noise reduction, when in fact no noise reduction takes place during loud passages. It isn't necessary...the music itself masks the noise, so noise reduction isn't necessary. That's one of the two ways that lossy compression techniques such as aac+ in HD Radio, mp3, dolby digital, wma, etc. work. They do two things. One...they allow noise to rise and fall when loud sounds are there to mask them. If that's not sufficient (to allow efficient coding of the audio with the available bits), they allow soft tones to be ignored in the presence of similarly pitched loud ones. This explains the weird "comb filter" effect on low bitrate digital audio, such as 20kbps web streams.
Well ALL loud audio masks ALL quiet audio to some degree. A spectrum analyzer shows that the vast majority of musical energy is below 2khz. Loud sound in the midrange (and bass) makes it impossible for even those of us blessed with the most golden of ears to tell whether anything is present above 20khz or not. This can be demonstrated with a/b/x comparison, and has been repeatedly.
There's another really bad thing about extending high frequency response too high. While there may not be any musical benefit, there could be real problems...such as high frequency noise from computer monitors (what studio doesn't have a bunch of those?), lighting, and all kinds of electrical and electronic doo-dads. There IS audio "up there". Just ask your dog. These "sounds" may be beyond the range of typical human hearing, but they can EASILY "beat with" sound IN the audio band (such as musical instruments and voices) to produce distortion components that are below 20khz...IN THE AUDIO BAND. Not only that, but recording such "sounds" may make a recording more "accurate" (after all they were present in the studio), but it also robs the recording of the very resolution that was gained by adding the extra bits and or increased sample rate. Bits are being wasted recording things YOU CAN'T HEAR, and that aren't even a parf of the recording. All good arguments that the "purist" approach is to NOT extend frequency response much beyond 20khz! Something to think about
