Well, I know the rule of dont dither twice therefore I usually come across this problem :
If your host doesn't support freeze then you'll probably need to bounce some VSTi to wave.
Normaly I'd do that by exporting 88khz 24bit, then downsample with Voxengo r8brain and import back as 44Khz 16bit.
Problem is that when eventually you gonna export the entire mix you're gonna do the same thing... export at 88khz 24bit and downsample to 44khz 16bit.
So you end up with the bounced parts getting dithered twice - once when you exported & imported them in order to save CPU usage and the 2nd time when you exported the entire mix.
What should I do in such scenarios ?
No dithering twice!
-
- KVRAF
- 12977 posts since 29 Sep, 2003 from Ottawa, Canada
Skip the r8brain part and just render to 88k wav?
Greg
Greg
-
- KVRAF
- 6478 posts since 16 Dec, 2002
in projects where you are working with 24bit files, you really don't need to worry about it. The noise will be so low that it just doesn't matter. If given the choice or forced to dither from 32bit vst pluginpath to 24bit, do a dither without noiseshaping (usually in fact gaussian). It's the safest and mathwise most transparent, you won't hear the dynamic loss. Especially when working with variable samplerates, never use the noiseshaped dithers, apart from the absolutely last step. Even then it doesn't really matter.
But when going to 16bits at any point, or working with 16 bit files. Only ever do 16bit dither as the last step. The worst possible thing to do would be to do more than one noiseshaped dither at 16bit modes. The added shaped noise might creep up and become audible.
Remember, dither=adding noise. But at its most optimal form preserving dynamic range.
With all this in mind, it should be clear that 24bit dither, even if repeated will have no (or negligible) impact on the much louder 16bit dither. It gets buried under.
But when going to 16bits at any point, or working with 16 bit files. Only ever do 16bit dither as the last step. The worst possible thing to do would be to do more than one noiseshaped dither at 16bit modes. The added shaped noise might creep up and become audible.
Remember, dither=adding noise. But at its most optimal form preserving dynamic range.
With all this in mind, it should be clear that 24bit dither, even if repeated will have no (or negligible) impact on the much louder 16bit dither. It gets buried under.
-
- KVRAF
- 8705 posts since 24 May, 2002 from Tutukaka, New Zealand
I've often had the impression that some people here dither every time they bounce...another reason for leaving things well alone unless you learn something about them.
Recently I've gone over to rendering everything at 44.1/32bit (although I used to record at 44.1/24)...that way I really don't have to worry about dithering until the end at all.
Having said that, with the sort of mixes I generally make, I'm not even sure I noticed any problems previously with no dithering at 16bit. But then I don't do particularly quiet and subtle dynamic music...no real instrumentation, I often don't even use velocity with synths (except to control filters), and I tend to apply reverbs etc during the final mix without rendering them earlier. So the jury is still out for me with dithering. If it means I can't save one song to a CD for storage, then I'll have to think long and hard about whether I really want to use dither at all.
If on the other hand, I were to send stuff out for mastering, I most definitely would not use dithering at all, although I'd record at 32bit.
I suspect that for the most of us doing your average loud electronic stuff, dithering is like the high samplerate argument....technically better, but ultimately undiscernable.
Recently I've gone over to rendering everything at 44.1/32bit (although I used to record at 44.1/24)...that way I really don't have to worry about dithering until the end at all.
Having said that, with the sort of mixes I generally make, I'm not even sure I noticed any problems previously with no dithering at 16bit. But then I don't do particularly quiet and subtle dynamic music...no real instrumentation, I often don't even use velocity with synths (except to control filters), and I tend to apply reverbs etc during the final mix without rendering them earlier. So the jury is still out for me with dithering. If it means I can't save one song to a CD for storage, then I'll have to think long and hard about whether I really want to use dither at all.
If on the other hand, I were to send stuff out for mastering, I most definitely would not use dithering at all, although I'd record at 32bit.
I suspect that for the most of us doing your average loud electronic stuff, dithering is like the high samplerate argument....technically better, but ultimately undiscernable.