Why would you dither a premaster? Im a master engineer and I would smack a man for doing that. There is never a need to dither twice.

This is very true!!!

monotremata wrote:Ok to clear up some Logic stuff...Youre audio is not at all 32 bits.No piece of hardware can top 24 anything youve recorded is at 24 bits and thats it. This is why there is only ONE checkbox in the Audio prefs that says 'Enable 24 bit recording' and you have no other options. Its either 16 or 24 depending on whether youve enabled it or not.That's just because Logic and the vast majority of sound cards don't support 32 bit recording. You're referring to the bit depth of the source audio files. If you mix or process those files in any way then you're creating numbers of a higher resolution, hence...Logic uses 32 bit floating point for its internal number crunching and processing.You're not losing anything when you bounce your 24 bit masters with it as your audio files arent being touched or truncated at all.Of course you are - you're losing all of the extra resolution generated by your DAW's mix bus which naturally runs at a far higher resolution. When you mix multiple audio files, your DAW is essentially taking numbers from each sample and summing them together. To avoid overloading your mix bus, you would likely have to pull the volume down. This is achieved by multiplying or dividing the samples.For sake of illustrating a point, if we pretend for a second that audio files were 2 bit and we took two samples 3 (11) and 2 (10) and mixed then we'd end up with 5 (101), a 3 bit number. If we divide that by two we end up with 2.5, which isn't exactly representable in 2 bit binary, so in that hypothetical scenario, either truncation (meaning the sample would be rounded to 10) or dither (meaning the sample could randomly end up on either 10 or 11) would have to be applied.The only audio files Logic actually stores at 32 bits are when you Freeze instrument tracks.The rest of it is all just processing power.Re: Processing power - you mean numbers. Doesn't really matter whether they are files on your hard-drive, in your RAM or being processed by your CPU.I wouldnt dither shit if I was sending it to a mastering house. Id just bounce my 24/48k tracks and send them those. If I were doing it myself, everything gets bounced out of Logic at 24/48, then imported into Waveburner.. CD layout gets done, any mastering plugs setup, then when you actually bounce it all to your final CD image, thats when the dithering gets done.So you would truncate, that's fine, but why? Why not dither to 24 bit since bit reduction is clearly occurring going from a 32 bit float mix bus to 24 bit file? My original question was whether the noise generated by an extra stage of dither at 24 bit was better or worse than the quantisation errors introduced by truncating this extra detail.

Ferrett wrote:Why would you dither a premaster? Im a master engineer and I would smack a man for doing that. There is never a need to dither twice.If we're going from a 32 bit float mix bus to 24 bit file then bit reduction is clearly occurring. It would surely make sense to dither it (at 24 bit of course) rather than truncate it if given the option. (Bear in mind I'm talking theoretically here, in reality it doesn't make any significant difference whether the file is dithered or not at 24 bit because the resolution is so high that it doesn't really matter either way)It sounds to me that you're following the only-dither-once rule at the expense of the dither-not-truncate rule, but you're not really saying why. Surely there is a need to dither twice (or even more!) times, *if* bit reduction is occurring multiple times?

You know, i'm still on the fence as to how much effect dithering has. I always think that one needs to compare the dithered and non-dithered material side by side in an A/B test. I tried it with some of my mixes in 32 bit floating point in Wavelab Essentials 6, and didn't see a noticeable difference by not doing it....there may have been, but I honestly didn't perceive a difference. The music was done in 16/44.1, so maybe it loses more when at 24 bit. I've recently started doing some stuff in 24/48, so i'll see what happens. I do notice that the recordings already have a bit more top end clarity and punch to them, and it will be interesting to hear what happens when they get translated back to 16/44.1.