Sample & bit rate

Question

I have a two part question:
1) I have a Presonus Firebox that allows me to record up to 24 bit/96k. I am wonder if I should take advantage of this and use those settings in audio midi setup for when I record into GrarageBand? Is it worth the extra file size? Is there a noticeable quality gain? Will 96k work with GB?
2) I know GB is only 16 bit and that CD's that it will be mixed to are at 16 bit/44.1 kHz. But what about if I record a song in GB at 24 bit/96k, mix down to iTunes, and put it on my ipod as an AIFF will the increased quality show up there? What about on an Audio DVD made from itunes with iDVD.
Thanks for any info and/or opinions,
Paul

Answer 1

Not that it's relevant in the context of GarageBand, but it might be useful to clarify the effect of sample rate and bit depth.
The sample rate affects the frequency response. it is mathematically impossible for a sampling system to correctly render any frequency at or above half the sampling rate (trying to do so produces false frequencies - 'birdies ' - known as aliasing). 44.1kHz effectively limits your frequency response to about 21 kHz, 48kHz to about 23 kHz, and so on.
The bit depth affects the noise floor: 16-bit gives a theoretical noise floor of about -90dB below peak: 24-bit gives a theoretical noise floor of about -138dB below peak. -90 is pretty well inaudible: but remember that when adding tracks together the noise, being random, is added (3dB per track if they are the same level when mixed - of course the peak level increases too). The digital noise, if it becomes audible, is far more annoying in its sound than good old-fashioned tape hiss.
In practice, GB's restriction to 44.1/16-bit isn't really that much of a problem: of course 96/24 would be better, but you gets what you paid for!
Incidentally, don't confuse bit depth with bit-rate, which is a combination of the sampling rate and the bit depth, and is fixed for uncompressed files: but of course as you compress files with MP3 or similar you reduce the bit-rate by leaving out information, even though the sampling rate is the same.