(I hope this is the right forum section.)

Foobar2k and Winamp have a "Use dither" option. On the Winamp forums one can read that this option is intended for 16bit files while using 16bit playback.

However, the article on dither in Wikipedia mentions that dither should be applied while converting to a different depth (16->24 and vice versa).

So can anyone enlighten me what the point is in dithering 16bit->16bit?

For dithering check http://en.wikipedia.org/wiki/Dither, specialy audio section.
In short there is no sense in dithering 16->16bits.

Do you mean playback of 16-bit lossless files, or lossy files (MP3/OGG/WMA etc)? For 16-bit lossless, no, dither makes no sense. For lossy, however, it can make a difference - theoretically, at least, and, under carefully controlled conditions (and with a very good set of ears), audibly.

Lossy files have no bit depth per se, regardless of the bit depth of the incoming PCM data - the encoded data is stored as floating point, and decoded to whatever fixed bit depth is asked for from the end user/decoder.

Thanks for the clarification.

(Btw, Winamp's "Use dither" option works only for Vorbis (*.ogg) and FLAC files; so in the case of FLAC it's really useless it seems...)

Generally, a proper dither should be applied after each operation on audio. So if you are using amplification (atenuation), for example, dithering should be enabled even for lossless files. It's useless if the player is making no modification to the audio (use kernel streaming for playback straght from decoder)

You slightly misunderstood the article. Dithering is needed to get rid of audible digital distortion which is caused by rounding/truncation during the process of lowering the bitdepth. The lower the bitdepth compared to the original signal is, the more erronous the result will be, making additional dithering necessary. But increasing bitdepths don't suffer from any rounding or truncation errors since they can always fully represent the original values, padded with additional zeros.

if you want to see/hear how dither works or sounds, I did a small program to add dither and listen to sound generator or wav files. It's called ditherer and it's free :

on http://dither123.dyndns.org you'll be able to hear the difference between difference...

A MP3 decoder routine internally works with higher resolution and outputs floating point variables (or 24 or even 32 bits integers) So to go back to 16 bits involves truncation/round off.

So Winamp's current behavior of not using dither for MP3 playback can result in sub-optimal results (from an "audiophile" point of view, save for audiophiles not using Winamp or MP3 in first place ).

Are there MP3 plugins to replace the Winamp default one that use dither? (I know of some replacements, but no idea if they dither or not).

The MAD plug-in features options for dithering and noise shaping. I don't know how well they work since I haven't tested them so far, due to my soundcard's output being set to 24 bits anyway. Rounding errors should be well below the noise floor of analogue devices in this case.

I'm using mpd in Linux for vorbis files. What mpd does is hand off
the vorbis stream and replaygain values to libvorbis and then give
the result to alsa-lib, which just pads everything to s32le and
routes it directly into my M2496.

I can't test if dithering sounds different, because mpd doesn't have
support for that. But just out of curiosity, do you think that
dithering would increase the output quality in this scenario?

Regards
mic

Yes, winamp 5.3 (and onwards) default mp3 decoder. (see Preferences->Playback)
(built-in feature of those versions, so it's useless to use the new plugin in an older release)

DJ Egg disagrees: http://forums.winamp.com/showthread.php?postid=2051799



I don't know what version that was, but there's nothing in the changelog.