From: Tim Ashworth
I have a simple question that I can’t find a concrete answer to and thought you may be able to shed some light. Should dither be applied when converting 24 bit lossless source (FLAC) to 16 bit lossy output (mp3)? Thanks if you can shed some light on this topic.
Normally, as long as wordlength is being reduced, DITHER MUST be applied. But mp3 (as well as AAC) is a special case because the LAME and Fraunhofer encoders work in 32 bit float in their encode side, and accept up to 32 bit words without penalty and without need for early reduction. They then use all the information you send them (24 bits in your case) to produce the best mp3 that they know how to produce, and internally take care of getting the most resolution encoded into that final 16 bit mp3 word. In fact, although the mp3 word is a 16-bit file, when playing the mp3 it will decode to full 32-bit float. It’s still lossy, but it’s not as lossy as you might think!
In other words, an mp3 made from a 24 bit or 32 bit float source file has the potential to sound better than one made from a 16-bit; you may perceive more depth or better resolution, if you use the highest bitrate mp3 you can make. You may not even hear a difference, however, if you make only a 128 k mp3…
Don’t forget to convert the lossless FLAC to a 24 bit wav first! For more information on this subject, I suggest you read my book: iTunes Music.