Just curious: why would you dither a mastered WAV from 32 to 24 bit before converting to MP3?
What's the advantage in doing that?

Well you should always dither when going from a higher bit depth to a lower one.
 
There is no interface capable of playing back a 32 bit file so in the example above, 24 bit must be his final master.
If it was for CD release you'd dither from 32 bit to 16 bit with a 44.1KHz sample rate, like I do

There seem to be some comments that confuse export bit depth with internal mixing word length. There is no audible difference to me between 24 bit and 32 bit export bit depth yet I always use the 32 bit option when exporting a mix to be mastered just in case.
 
Always dither when going for a shorter bit length ... never dither otherwise.
 
Always use the 64 bit internal world length if possible. That does make a difference.

Jyri T.
Always use the 64 bit internal world length if possible. That does make a difference.

If you include an "imaginary difference" as a difference. Most of the "32 bit errors" everyone worries about don't ever make it to the 24 bit output and simply aren't audible in the real world regardless.

drewfx1

Jyri T.
Always use the 64 bit internal world length if possible. That does make a difference.

If you include an "imaginary difference" as a difference. Most of the "32 bit errors" everyone worries about don't ever make it to the 24 bit output and simply aren't audible in the real world regardless.

I beg to disagree --- if you have a hefty track count ant a ton of fx. It adds up at some point so that there is a difference.
If you have a reasonably small amount of tracks and effects, you're right.

Jyri T.

drewfx1

Jyri T.
Always use the 64 bit internal world length if possible. That does make a difference.

If you include an "imaginary difference" as a difference. Most of the "32 bit errors" everyone worries about don't ever make it to the 24 bit output and simply aren't audible in the real world regardless.

I beg to disagree --- if you have a hefty track count ant a ton of fx. It adds up at some point so that there is a difference.
If you have a reasonably small amount of tracks and effects, you're right.

You are disagreeing based on what?
 
The reality is the errors simply don't accumulate to the point of being any kind of issue in the real world. We can go through it in detail if need be.