It seems that Lame 3.90.3 is the most popular encoder,
and --alt-preset standard, the most popular switch, but I
haven't seen alot about decoders.

What are folks thoughts on the best playback encoder, and
which software uses what?

Are the playback decoders less/more important than the
encoder?

Can most/any popular players select from different playback
decoders?

Thanks all, Dex

foobar2000 has the best decoder.

edit: typo

Which mp3 decoder is used in foobar?

Exactly . . . that was the question.

Dex

Well, it may not be the best, but the developer's always doing the Right Thing™ when bugs are found, so if it isn't presently, it will be sooner or later.

fb2k uses a modified version of the mpg123 decoding engine, IIRC. But there are so many patches and bug fixes that they are no longer really similar at all.

Winamp's decoder's made by FhG, which is sort of the defacto standard, seeing as how FhG is one of the developers of the MP3 format.

And, of course, there are always the players based on BASS.DLL, like XMPlay (not exactly accurate, I know, but they're by the same dev, so I'm assuming they're based off of each other). BASS has been developed for a very long time, and is excellent at playing modules. Seeing as how that's rather difficult, I'd assume it would be a competent MP3 decoder as well.

Really, the only difference between any of them is the way they deal with errors. Or that's all it should be. It is far, far less important than the encoder step. Most are bit-identical with dithering turned off. The only difference is in post-processing, really, although there are certainly some non-compliant decoders.

Most players only use one decoder, though, and bug-fix it as necessary.

foobar2000 uses a heavily bugfixed/modified version of mpglib, which is also used for lame's decoding capabilities.
BASS.dll AFAIK has no MP3 decoding capabilites at all.

The general consensus is that MAD, mpglib and FhG's are the most compliant/perfect MP3 decoders. Most commercial software nowadays will probably/hopefully use FhG's.

dev0

out of date, but I can't believe no one linked to it (the mpg123 based decoders did well in this test):

www.mp3decoders.org mp3 decoder quality test

Cheers,
David.

EDIT: mpg123 test "results"

Oh!

I always thought MAD was the best mp3 decoder, and I also thought foobar used MAD!

I woulda, but I was too lazy and pressed for time to google for the link. Thanks.

Not true.

MAD is GPLd, it can't be used on Foobar unless Peter make it open source.

I guess, part of what I'm asking is: with encoders, its clear what all the names are. I can use AudioGrabber, EAC, CDex, etc, and select whatever encoder I want, ie., Lame, Xing, Blade, etc. But for some reason, the names of decoders seems a bit more elusive.

I've heard of MAD, but prior to reading the responses in this thread, I don't think I've heard of any others. I'm just wondering why that is, which also leads to other questions, like: I can get a MAD plug-in for Winamp. Is it better than whatever Winamp uses now?

Its those types of questions. If I'm going to go to all the trouble of making mp3's as best I can, it would be nice to play them back with highest possible fidelity potential also.

Thanks for the reply's so far folks. Dex

MAD is dithered. The default decoder is not. Whether or not dither translates into better audio quality is presently up to debate. The tests for dither/non-dither are presently done at 8-bit, not 16-bit, merely because the loss is more apparent at 8-bit, so although you can typically infer that dither algorithms that are better at 8-bit are better at 16-bit, the few tests that have shown a difference were ambivalent as to whether or not dither was beneficial to the final output signal. At 8-bit, dither is preferable to non-dither. It's an absurdly low noise floor anyhow, so it shouldn't make that much difference.

Furthermore, the audio is sampled at 16-bit, so information below that noise floor is probably useless anyhow.

In other words, compare the two yourself. There are at present no meaningful tests that show the superiority of dithered 16-bit output to non-dithered 16-bit output. If you find you prefer one, stick with it.

PS. I wrote this while under the influence (against what is probably my better judgement, but then again, that should be impaired), so take most of this with a grain of salt. I'll check this in the morning to make sure I'm not spewing nonsense.

Actually, that was quite informative and well articulated.

By ". . . take it with a grain of salt", I assume we're doing Margaritas?

Dex