I record at 24/44.1 if the eventual requirement is audio-only, and 24/48 if my sound track will be matched to video. Then I deliver whatever is specifically requested, whether it be mp3 files, CD-Rs, etc.
--aaronji wrote:
> A major advantage, as I understand it, is that the extra bits greatly reduce the noise introduced from quantization error during analog to digital conversion (each extra bit approximately halves the noise). So even if the recording peaks considerably below full-scale, you can amplify it without making the quantization noise audible...
Aaronji, that just isn't true as a rule. It all depends on the noise in the signal that's being quantized versus the converter's own noise floor. The quantization noise of the converter only needs to be a few dB lower than the noise in the incoming analog signal, however many bits that takes. Any further bits beyond that will (or should!) be random; therefore they don't contribute to the precision or accuracy of the recording in any way. The notion that computation (post-processing) will somehow work out more accurately with more bits is a total fallacy when those bits contain only noise.
Of course if they contain meaningful signal information, then by all means keep them--but from what I've read on this board over the years, I think that a lot of people here are kind of fooling themselves on that point. For live classical recording, which is what I mostly do, the signal going into the a/d converter might have a noise floor maybe 60 or (if I'm very lucky) 65 dB below the peak levels. 14-bit quantization can handle that (keeping in mind that the converter noise needs to be below the incoming noise at all frequencies). 16-bit recording allows for more conservative level settings and fewer accidental overloads--plus it's been decades since 14-bit recording was even an option on available recorders. But for example the BBC used 10-bit (analog companded) studio-to-transmitter links for their classical broadcasts for years, and they were justly renowned for their audio quality.
24-bit recording (which might have 18 or 19 actual bits to offer, when the noise level of the converters and the rest of the signal chain is taken into account) is even nicer than 16, but it's clearly a luxury for our benefit as live engineers. I use it because it doesn't hurt anything*, and once in a blue moon it helps a little. But I've never delivered a 24-bit recording to a client; people not directly involved in recording have no use for the extra data, which only causes them trouble and expense.
(Notes: If the extra, unneeded converter bits beyond the incoming signal's noise floor aren't random, then they're adding distortion to the recording, and any such converter should be removed from service immediately. And please keep in mind that the noise floor of a converter is whatever its ACTUAL value is, with the number of bits setting a theoretical limit that is never reached in practice; no physically realizable 24-bit converter has an analog input noise floor anywhere near 144 dB below peak level. So please don't make a mental leap from "24 bits" to "8 bits quieter than a 16-bit converter" because it's much more likely in reality to be only 2 or 3 bits quieter, if that.)
--best regards
_________
* Edited later to add an asterisk above, and an important qualifying remark as a footnote: 24-bit recording "doesn't hurt anything" ONLY if you set your recording levels appropriately. At the risk of seeming to be unkind, let me say this as clearly as I can: If you're one of those people who completely misunderstand the concept of headroom, who actually AIM to achieve peak levels of -12 dB or even lower (instead of aiming to get your peak levels as close as possible to 0 dB without actually hitting it), then the use of 24-bit recording is hurting you by "enabling" your basic failure to understand the concept of recording levels, and your "24-bit" recordings are quite possibly noisier than they would be if you learned how digital recording actually works, and used 16-bit recording appropriately.