The 64 bit aspect of all this again just applies to steps 2 and 3 above. Sonar a while acquired an option to set the internal mixing system to "double precision 64 bit". I have that switched on, but I couldn't honestly claim I can reliably tell a difference. But all else remains the same as above.

Sonar 5 does have the 64-bit precision mix engine.  I have that turned on.  But the maximum file bit depth it allows to set/export at is 32 bit.  That's just yet another aspect of what is confusing me.

Ah, right. Yeah, there's a lot of shared terminology among different contexts.

SonicExplorer
Sonar 5 does have the 64-bit precision mix engine.  I have that turned on.  But the maximum file bit depth it allows to set/export at is 32 bit.  That's just yet another aspect of what is confusing me.

The 64 bit part is simply referring the the precision of the maths involved when performing any action on an audio file - it has NOTHING to do with the bit depth of files which as you have found, you can export as 32 bit.
You'll struggle to find a player that can actually play a 32 bit file, there are several that can handle 24 bit files but CD of course is limited to 16 bit

There's a good potential spin off topic from this, about what does and doesn't make significant and decisive differences in the real-world deployment of music. Should probably be its own thread, mind you.

Hehe, good thing you mentioned that, since as I was reading this during lunch "frequency masking" (where one loses audio tracking capability even if the smallest detail were present in the masked frequency) and the fact that much deployed music is played as an mp3 through ear buds came to mind (not to mention you cannot control the EQ settings of the listener's mixer). So many variables to the "system" that arguing them one-by-one is moot.
 
Back to the OP's concern/question... So much of a song falls into "musical content" versus "technical content," so don't get too wrapped around the axle on things "because you can." John's 5 step summary above is very to-the-point about each phase.

Bristol_Jonesey

SonicExplorer
Sonar 5 does have the 64-bit precision mix engine.  I have that turned on.  But the maximum file bit depth it allows to set/export at is 32 bit.  That's just yet another aspect of what is confusing me.

The 64 bit part is simply referring the the precision of the maths involved when performing any action on an audio file - it has NOTHING to do with the bit depth of files which as you have found, you can export as 32 bit.

"Nothing to do" is a bit confusing in this context. Sure these bits are related. When the engine is working in 64bit mode, it produce 64bit floating point numbers. These can be saved "as is" into 64bit format file or can be "truncated" to 32 bits (or converted into 16 or 24 bit integers).
 
So, nothing is wrong with writing 64bit files. The whole discussion here is about usefulness to do this. And in that aspect everyone except one (self claimed) "expert" agree that exporting into 64bit files make no sense.
 
To prevent the question why everyone recommend to mix with 64bit engine but say more then a half of it is a "garbage". During mixing you normally use some plug-ins. F.e. in VST2 standard there are 2 methods plug-ins can implement, one accept "float" (32bit) data and another accept "double" (64bit) data. And if some developer has decided to use "floats" for intermediate calculation, the quality of the result can degrade.
 
Let say we have 3 significant (decimal) digits precision with input "123". We use an algorithm which adds 5000 and then deducts 5010 (5000=5*10^3,  1 digit precision; 5010=501*10, 3 digits precision; so all 3 used numbers have at most 3 digits precision) :
1) if we have unlimited (or at least 4) digits DURING the calculating, the result will be:
   123+5000 = 5123, 5123 - 5010 = 113.
2) if we have only 3 significant digits DURING the calculation, the result will be:
   123+5000 = 5120 (the last digit is truncated!), 5120 - 5010 = 110.
Note that in (1)  INPUT and the OUTPUT have 3 digits precision. But in (2) we have LOST the precision, we get 2 instead of 3. So in this case we do not need more then 3 for "files", but we need at least 4 for "processing"
 
Sorry for a bit "scientific" and for sure abstract example. But unlike in medicine, military and atom reactor calculations, I do not think that many "creative" developers of "vintage analog gears emulations" calculate used algorithms precision.  I guess most of them simply do not know how to do this and many of them even not aware the problem exists. Also in the SoftSynth world, where "creative" user can turn any from 2000 parameters to extreme values during preset creation, it is impossible to keep calculation under control.

https://myl8test.files.wordpress.com/2013/05/download-es9018-datasheet-here.pdf
 
jitter control etc looks a bit like the phase accumulator asic I designed back in 1986...
 

Post deleted

Since this topic hasn't 'died, I've found some resources on the topic of 64bit fp audio files. Whilst these aren't science or digital audio theory papers, the writer seems to be echoing everything I've stated.

The first article, https://theproaudiofiles.com/6-db-headroom-mastering-myth-explained/ the writer suggests that there are benefits to exporting to a 64bitfp pre-master even if that pre-master is going to a 'pro' mastering house and will be eventually truncated ;) (Sorry to the poster above, yes I should use correct terminology) to 24bit and 16bit. The writer of the article also backs up my belief that 64bit fp audio files are container files and even if you record at 16 or 24 bit, because once you apply processing, the file becomes a defacto 64bit fp audio file :). So if you don't want truncation to occur it's important to keep the bit depth of your audio files as high as possible for as long as possible.

This article, www.justmastering.com/article-mixingformp3.php suggests that you should use a lossy format to upload to soundcloud. Whilst it doesn't suggest 64bitfp audio files - since 24bit or 16bit audio files are lossy files when converting from 64bitfp audio files - it follows that 64bitfp would be the best format used for soundcloud and possibly other streaming services.

This article is interesting www.justmastering.com/article-masteredforitunes.php because it says even if you hand in a 24bit audio file, during transcoding the file gets shunted back up to a 32bitfp audio file, so take from that what you want. If you read on, the article even suggests the higher the bit depth, including floating point audio, that you submit your master the better.

I had another article extolling the virtues of high bitrate masters for video, but I can't find that at the moment. And even though the writers of the above article haven't used scientific data, and therefore I won't say that I'm right full stop...I believe these articles do seem to back up my belief in 64bitfp audio files for not only processing but also for masters as well - no matter what the final format of said 64bitfp masters will be.

There you go John T

Peace