View Single Post
Old 26th June 2005
If you're going to mix "outside the box" (ie, send out submixes through a multichannel i/o) and you have lots of computer horsepower, you might want to experiment with higher sampling rates.

But if you keep your production chain in the digital domain from tracking to CD or mp3, you'll probably want to keep your work at 44.1 kHz -- or an even multiple of it like 88.2 or 176.4 -- to avoid the signal deterioration associated with downsampling from a rate that is an uneven multiple of the target rate.

[Now, if you're outputting to DVD for video or 48 or 96 kHz DVD audio, OTOH, and staying in the digital domain, it's certainly best to work at 48 kHz or an even multiple, 96 or 192 kHz.]

While it makes perfect sense to work at 24 bits (or higher) even if you're outputting to 16 bits for CD or mp3s, because of the uneven-multiple-downsample issue, most of the folks I respect work at their target sample rate -- and if they want to maximize their fidelity during the production stage or have a high-def copy for archival, they work at double the sample rate.

So, you know, the short answer is, it's nice to have high resolution as an option should you ever need it -- but for most folks, who need to balance computer horsepower requirements with plug-in use and track count, most folks find working at higher dynamic resolution (bit depth) to be far more important.*

* -- since each added bit doubles the resolution... a 16 bit binary number can hold any of roughly 64,000 values. A 24 bit number can hold any of 16 million values. That's a 256 fold increase in dynamic resolution for only a 50% increase in processing overhead. In most cases, any reasonably trained ear can easily hear the difference between 16 bit sound and 24 bit sound. But few subjects in tests of 48 kHz versus 96 kHz at the same bit depth have been able to consistently tell the diff. Additionally, blindfold tests conducted (I think) by Apogee in the 90's suggested that most subjects preferred 20 bit 48 kHz over 16 bit 96 kHz, further corroborating the generally accepted wisdom that bit depth increase trumps sample rate increase at the typical rates in use in the last 10 years. [Coffee. It's the coffee talking.]