Depends what your final product is to be. If you are working, as you say, in 44/32, and you want to make a CD, you MUST make 44/16 files. You can do this without dithering, but it is not wise.
If you are doing work for a client and they want 44/24, same thing. You can do this without dithering, but it is not wise.
Along with Jeff’s reference, here are a couple more threads on this topic.
Here is my take:
_For me it all depends on the type of music.
I can notice a difference if I am recording a single flute/ slack key guitar or piano or violin etc.
Also I have also recorded string quartets and (light/smooth) jazz trios and I can notice even then.
But If I am recording a pop or reggae or rock band or even a ‘full out jazz trio’ etc I can’t really ‘hear’ the difference.
IMHO
it best to use your ears and see if the dithering noise
sounds better than the artifacts._
Now, I generally record at 44/32. I have an album I am in the middle of mixing at the moment and my mastering engineer can receive files at 44/24. So when I render my mixes I need to dither down from 32 float to 24 bit, right?
(My gut says yes but I thought I’d toss this out to geeks that are much smarter about such things than ‘my gut’).
So when I render my mixes I need to dither down from 32 float to 24 bit, right?
I would just ask the mastering engineer. Depending on his/her gear, it might be able
to handle yer 32 bit files.
Suggestion:
Since you don’t already know this info, it sounds to me like this is the 1st time
dealing with this person.
If you plan to work with a diff mastering engineer in the future, perhaps it would be best to
find out what is needed by them BEFORE you start the next recording.
That way you can record at a level that completely matches what they need.
Curt,
I have been in discussions this morning with my Mastering Guru. Here is his reply to my question:
"Dithering is always a good idea when reducing ordlength, even if the difference will be @ -140dBSPL, imo. Can’t hurt.
That said, I’d NOT use UV22HR - noise shaped dithers are a bad idea when further processing is likely to happen - just flat TPDF dither is ideal."
He’s got a couple of Grammy’s to his credit so I trust his judgment.
So this leads me to my next question. Do we have access in Cubase to Dithering OTHER than UV22? Is there something else under the hood in Cubase 7.5?
As long as your 32bit floating point files never go full-scale digitally, then there is NO need to dither before giving the files to the ME, as those 32bit files are only 24bit resolution.
In fact the further any 32bit float sample is from full-scale, the more accurate it is compared to 24bit integer ones. This is because the latter are absolute, so the closer their samples are to zero, the less resolution they have, whereas the 32bit floats ALWAYS have 24bit resolution (though the lower bits may have been cut off during previous processing!).
I don’t know what the reason is (bit depth or better processing), but I found that the 24bit samples of Superior Drummer 2.0 sound better than the 16bit ones of EZDrummer, with the latter sounding rather noticeably flatter in comparison.