16 and 32 bit dithering
Oh Wise One:
If I’m taking a track which was recorded in 16 bits and making it into an audio file to place on a mastering track, do I save it in the audio file as 16 or 32 bits? Can I save it as a 32 bit file so long as I dither it down to 16 for placement on a CD or does bumping a track up to 32 bits destroy any of its audio integrity?
I don’t know what you mean when you say “to place on a mastering track”???
Do you mean that you are putting files on a data disc (CD-ROM or DVD-ROM or portable drive) to send to a mastering engineer?
Or do you mean loading the file into a track in some software program in which you are going to attempt to do the mastering yourself?
If it’s the first case, I’d just leave it alone and let the mastering engineer do whatever is needed for the way in which he works. You don’t want to screw it up and damage the audio quality in any way. Of course, mastering guys usually prefer 24 bit files and you should try to mix all your songs to 24 bit fixed point format, since that is somewhat of an accepted standard these days. But, if the file you want to master is already at 16 bits, then just leave it alone and send it as is.
If it’s the second case, then it depends on the software you are using, and how you plan on working with the file. If you are going to load the file into a track in a DAW program, such as Cubase, or in an editing program like Wavelab, and then are just going to use track insert plug-ins, or master section plug-ins, then you don’t have to worry about doing any bit depth conversions yourself. Most of these types of software programs these days have 32 bit floating point audio processing engines in them and all the plug-in operations and such will be done at full 32 bit floating point resolution, unless the plug-in itself is a fixed point plug-in… either way, the result coming out of the audio bus will be 32 bit floating point, regardless of what the original bit depth of the file was. Then, all you need to do is to save the resultant file at 24 bit fixed point, or 32 bit floating point, when you render the final results to a new file. When working this way, you don’t gain anything at all by converting the file before you import it to a track as long as you are doing non-destructive processing and rendering the results to a new file of higher resolution.
Now, on the other hand, if you are doing destructive plug-in processing where you apply a process and overwrite the original file with the results, then you would want to convert the file to a higher resolution format first. This would only happen for non-realtime processes that you select from the “process” menu in things like Wavelab, or SoundForge, or a DAW program, where it works directly on the file. This is usually NOT the preferred way to work… it’s better to use a realtime processing chain that keeps all the results at the full internal resolution of the program until the very final stage where you render the results down to the final format. In other words, you don’t want to apply one process at a time and save the results to a lower resolution format after each operation.
If you are doing this as your own self-mastering and need the final file to be 16 bits for a CD, then the last plug-in in your processing chain should be a dither plug-in set to 16 bits, and you would render the final file at 16 bits.
Bumping a file up to a higher resolution will not destroy anything… the extra bits will just be automatically filled with zeros until you do some process on the file that results in more than 16 bits of data.