Howdy folks,
 
I am trying to get my 64-bit 96k audio file out of SPLAT to master in Wavelab & make AAC files.
 
Wavelab will only accept 24-bit files.
 
I need to add dither after SRC & then truncate the bit depth, (all of which I want to do in Wavelab)
 
My understanding is that 64-bit is just a computational thing & that the audio will always be 24-bit. 
 
So, If I truncate the 64-bit Audio file out of the Hardware outputs at 24 bits am I losing aything? 
 
It is commonly said that you always add dither before reducing word length.
 
Am I reducing word length, going from 64-bit to 24-bit?
 
As you can see I am very confused. 
 
 

Yes, going from 64- to 24-bits is wordlength reduction. Dither basically takes care of rounding errors when you do that. It has nothing to do with SRC, which you don't need to perform anyway. 
 
When you export from SONAR, select 24 bits and Pow-R dithering and go. Don't worry about the 40 bits you're discarding.

Thank you for your speedy response bitflipper.
 
If I am making a 16-bit 44.1k file, I usually render mastering then convert sample rate from 96k to 44.1k, then add dither then truncate to 16 bit. This has always given me the best results. 
 
If I am going to export a dithered file to later add more dither, isn't that going to sound horrific?
 
Would I be better off just starting my sessions in 24 bit then I can transfer that file into Wavelab to render mastering, convert sample rate, add dither then reduce wordlength?
 

Yes, conventional wisdom is to only dither once.
 
Your 64 bit file, to my limited knowledge, cannot truly exist as an audio file (I don't know of any interface that would play it). As you say, the internal math computations are happening at 64 bit but the resulting audio file should be 24 bit.

Bristol_Jonesey
Your 64 bit file, to my limited knowledge, cannot truly exist as an audio file (I don't know of any interface that would play it). 

Of course it can - as with any audio file if the SW can read that file format it just converts it to the format the audio interface is set to. 
 
In terms of dithering down to 24 bit, it really doesn't matter as the dither noise in this case is of so low a level as to be irrelevant for any real world scenarios.
 
The rule to "always dither when reducing bit depth" is because if you don't dither you will lose any information below the noise floor when truncating. But at 24 bit the lowest bits are always buried in noise anyway, not to mention that they almost certainly contain nothing but noise in your signal anyway.

Are we sure we are talking word length here and not the processing 64 bit? 
Sonars default for processing is 32 bit and the option of upping that to 64 bit. 
But recording bit depth for every interface I read about is either 16 or 24 bit. 
What interface do you own that has a 64 bit recording word length? And if so what is gained buy such a crazy setting?? 0.00009%?? 
 

Cactus Music
Are we sure we are talking word length here and not the processing 64 bit? 
Sonars default for processing is 32 bit and the option of upping that to 64 bit. 
But recording bit depth for every interface I read about is either 16 or 24 bit. 
What interface do you own that has a 64 bit recording word length? And if so what is gained buy such a crazy setting?? 0.00009%?? 
 

Doesn't exist, hence my comment

That the interface can't record or playback 64 bit has nothing to do with it. You could save an audio file to 128 bit or 1024 bit for the purpose of editing, this will (at least in theory) prevent rounding errors when you use fx or bounce the file or change gain or any such thing. Upon playback of course the interface will play it at 24 bit but that doesn't mean the audio file "can't exist" at 64 bit. In fact the standard bit depth is 32 in Sonar I believe.

I figured 64-bit acted like 32-bit float point but the audio files exported from Sonar have a bit depth of 64.
 
Adding dither before truncating to 24 bit to use across different formats that will need further dithering worries me greatly.
 
Oh, I wish to understand computational science.
 
Thank you to everybody for the input so far.

Dithering to 24 bit really doesn't matter:
 
1. First of all, stuff at the level of 24 bit dither isn't remotely close to being audible under any real world conditions whatsoever.
 
2. Even if it was, the 24 bit dither would be buried under other noise - the best 24 bit converters can do is equivalent to 20 bits or so, so the dither is always buried under the noise from the converters in addition to other noise and it's all below what your ears can hear anyway.
 
3. Your audio almost certainly already contains nothing but noise at the same level as the 24 bit dither, which means you have "self dither" and dithering thus provides no benefit because the noise in your signal is already doing the dither's job.
 
4. You don't need to worry about repeatedly dithering causing noise buildup, because of #1 and #2 and the fact that you'd have to apply 24 bit dither literally dozens of times before it could possibly matter.
 
 
IOW 24 bit dither really just doesn't matter at all. You can dither or not dither or dither repeatedly at that 24 bit level and it will make no difference whatsoever.
 
You can think of it as changing the font on a street sign 2 miles away to make it "more legible" - it might make a difference from a distance that's just at the limits of your vision (like dither with 16 bit audio), but from miles away it's beyond irrelevant.