Export bits, dither, and MP3 encoding - thoughts?

My understanding (perhaps faulty) is that MP3 is inherently a 16-bit format, so bit reduction must be happening at the point of encoding if the input is higher than 16-bit. In that case, it might be better to do that with down-sampling and dithering algorithms of known high quality, rather than leaving it up the MP3 encoder.
 
That said, I usually export directly to MP3 from a 32-bit "Master Bounce" within SONAR, which begs the question: when exporting directly from SONAR to MP3 does SONAR do the bit-reduction and apply dither or does it just pass the raw 32-bit data directly to the MP3 encoder?
 
 
 
 
 
 

Seems like if you try to bounce a 24/32 bit track to mp3 manualy as opposed to the "What you Hear" option, Sonor won't let you. You have to convert to 16 bit.  Been a while since I ran into that, probably Sonor 6. May be different in 8.5.

I use the lame encoder and have it configured to work from within Sonar. It appears (to me) that Sonar produces a temporary dithered 16 bit file that is then encoded by lame.

I'm afraid my ears aren't good enough to tell whether a file was dithered by Sonar or the MP3 converter.

pdarg


But what about an exported mix which is then encoded to MP3 by another application (I am using WaveLab).

It seems to encode a 32 bit file just fine.

I probably work a bit differently than some around here.   There are two situations I generally deal with:
 
1- Exporting tracks and stems.   This is easy.  NO dithering, no changing from 24bit 44.1k in this case.
2- Creating a mix (stereo wav) and converting it to MP3.
 
For #2, I do my mix (bounce to track) in the project.  Then I use the Tool menu to send that mixdown to Adobe Audition.   In Audition it is loaded as a 32bit float.    When I'm done doing any tweaks (fades, whatever), I save the wav as a master mix (in the project folder, so everything is there).   Then I use the SAVE COPY AS in Audition to create the MP3 which I've pre-set to be the proper settings for a standard mp3.
 
Audition takes care of the rest.   It's not complicated and it gives me greater control.
 
You could probably do the same with WavLab.
 
 

bitflipper

I export from SONAR as a 32-bit file

This got me thinking ... my projects are always 24 bit, but the effects process in 32 bit float. Also, I've noticed when I freeze a 24 bit track, then unfreeze it, it has been converted to 32 bit.

I'm guessing this is why it would be good to export to 32 bit even though you are working with a 24 bit project? I normally have dither turned off in options so I should always export at at least 32 bit or am I thinking wrong?