I do multitrack recording on a Roland VS2480 in a small home studio. Everything is 24-bit.
What I want to do is transfer my 24-bit master to the PC, and use Audacity to bring down a few transient peaks, an
d raise the overall volume just a small amount.
Then, I want to dither that to 16-bit for the final master.
I’ve got the MBox2 with Pro Tools LE on 14-day trial right now, but I just found Audacity and love the simplicity
not to mention the price.
I love the Envelope shaper for handling the transient peaks, the lookahead limiter looks like it does a great job
also.
But, as a professional musician, I worry about losing ANYTHING from my 24-bit recordings that I spent so much time
on.
I’m assuming that everything is just 1s and 0s at this point. But how about internal rounding of the numbers, etc.
Is Audacity introducing any coarse rounding of numbers, or anything that might rob the original quality of the rec
ording when I perfom things like Amplify or the Envelope Shaper?
For the best sound quality, set the bit depth in Audacity to 32 bit (float).
“Edit > Preferences > Quality”
This does not change 24 bit audio data at all, but adds extra bits to reduce rounding off errors to virtually nothing.
Audacity will by default take care of dithering when you convert to 16 bit. You may also want to keep a 32 bit version in case you want to do further editing at a later date.
My preferred format is 32 bit 48kHz, then change to 16 bit 44.1kHz for Export to WAV ready for burning to CD.
There is also the problem that under certain circumstances, Audacity doesn’t deal with 24-bit sound very well. 32-bit floating, while not strictly a “universal standard” format, is very high quality and does translate into everything else really well.
<<<My preferred format is 32 bit 48kHz, then change to 16 bit 44.1kHz for Export to WAV ready for burning to CD.>>>
That’s 3 different places that errors could be introduced.
The first thing to worry about is the 24-bit master → computer transfer. Are you going to do this using either a digital S/PDIF signal or a digital file transfer (if that’s possible)? Or are you going to plug the analog output of your Roland into an analog input on your computer? If you go the analog route, then your sound card will undoubtedly add some amount of (probably inaudible) noise to the signal.
The Audacity edits are not worth worrying about, provided you don’t use the tools to mangle the audio. It sounds like you know what you’re doing though.
Steve suggested converting the 24-bit files to 32-bit in order to make your minor edits, but I don’t agree. The only real advantage to using 32-bit float is in a project that involves mixing quite a few tracks together. Since you’re working with only one track I see no reason to go through the format conversion. Doing so won’t lower your noise floor. Any possible advantage would only show up in making the Audacity edits more accurate, but all of that extra accuracy is only going to show up well below -144dB (the dynamic range of 24-bits). Since humans can’t safely listen to a signal with more than 120dB of dynamic range, you’re just wasting disc space by converting to 32-bit float. I say stick with 24-bit to make your edits.
Finally, the 24 → 16-bit conversion. This is obviously going to reduce quality, but there’s no way around that. By raising the overall volume level to peak as high as possible, you’ll minimize the difference in the noise floors. It sounds like that’s exactly what you’re trying to do.
Audacity automatically dithers as it outputs to a lower bit-depth format. You can even select the dither waveform in Audacity edit → preferences → quality. Here you can change the settings for the high-quality (aka export) dithering and interpolation. In the event that you’re not changing the sample rate, Audacity won’t interpolate the output file so that setting will be irrelevant.
Dithering has a lot of black magic behind it, but for your purpose the dithering waveform you want is called “Shaped” in Audacity.
As far as I know, converting from 24 bit to 32 bit should be lossless, or if not, then certainly too low to be audible.
Losses through A/D D/A conversion as the audio is transferred to the PC can be avoided by burning a data CD with the 24 bit data (with error checking).
Musicians can (and frequently are) VERY fussy about putting the finishing touches to their creations, and even with just 2 tracks, this can end up being dozens of small edits, minor tweaks to the Eq and dynamics, just getting rid of that slight whistling sound that nobody else even noticed…
Also he specified “as a professional musician, I worry about losing ANYTHING from my 24-bit recordings that I spent so much time on.” (note the capitalization). Whether or not the difference is audible (certainly won’t be for 90% of the population), 24 bit (source) > 32 (processing) > 16 bit (CD) will minimize losses. Audacity will also, in some circumstances, apply dithering during processing of audio data that is not 32 bit. As a one-off, this would be insignificant, but the question here was aimed at pristine quality over multiple operations.
Ill stick to my guns on this one alatham, we’ll have to agree to differ