23rd January 2001 - Final major site update, adding 14 new decoders, taking the total to 42. Added new tests for 24 bit accuracy, ID3v2 compatibility, Corrupt file decoding, and Free Format compatibility. Changed the layout of some tables. Removed references to "latest" Winamp being bad, since the latest Winamp (v2.7 at the time of writing) decodes perfectly.
27th October 2000 - No site updates due to heavy workload BUT alpha test version of new Winamp decoder is free of 100Hz bug [8191 bugfix] from version alpha13 onwards. See Christophe's .plan for information on the latest alpha test release.
27th October 2000 - 100Hz bug Winamp/lame problem explained. See here and here. (Huffman coding is a lossless compression technique, used in mp3 encoding to compress the data produced by the lossy psychoacoustic processing. Lame was allowing larger values than other codecs, which were not decoded properly by the Winamp decoder.)
7th September 2000 - Uploaded a new set of tests on ten more decoders, including CD writing packages. No CD writing package decodes mp3s perfectly! See list of decoders for the individual results.
I'm carrying out a PhD at the University of Essex, in Colchester, UK. My funding comes from the EPSRC. I have no affiliation with any of the companies whose products I have tested. I carried out this test as part of my own research, and have made the results available on line as they may be of interest to others.
I have no plans to test any further decoders. However, if anyone has any suggestions I'll be happy to listen, or explain how you can carry out the tests for yourself. If any decoder manufacturer wants me to verify their decoder, they should contact me. I'm happy to carry out private independent tests, or to publish the results of my tests. Anyone who thinks bribing or hassling me will be easier than fixing their faulty decoder will be politely refused.
This question is answered on another website, though this advice is somewhat out of date.
Some newer CD writing programs will decode on-the-fly, so you don't need to decode from .mp3 to .wav first - the CD writing program does it for you as it writes the CD.
| Advantages: | Disadvantages: |
|
|
If the decoder is working correctly, the audio quality will be the same whether the output is going to your sound card, a .wav file, or directly to a CD. The pros and cons of various CD writing packages with "on-the-fly" decoding capabilities are discussed in the Conclusion.
EXTRA NOTES: The most common problem when trying to make a CD that will play in a normal CD player (eg in the car or your hi-fi) is writing in data mode instead of audio mode. You can fit 10 CDs worth of mp3 files on 1 data CD, but no CD player will play it. You can even write 1 CDs worth of .wav files to a data CD, and again, no CD player will play it! To solve this problem, you don't need special media, but you do need to use the audio CD mode of your CD writing software. Easy CD Creator, Nero, WinOnCD etc will all do the job. DirectCD will not, as it can only make data CDs.
CD-RW discs will not play on most audio CD players - use CD-R blanks only.
CD-RW discs will not play on most audio CD players - use CD-R blanks only.
Please go back and read the Introduction. Encoders can and do vary. A decoder should do just what the encoder instructs it to do via the mp3 file. If it doesn't, it's faulty!
There are several reasons, including...
- Different wave file sizes do not necessarily mean that the wave files contain different audio data. The header of a wave file can contain additional data, and the decoder may or may not use this space. The decoder may also add extra silence to the end of the file. Some decoders recognise encoder added silence, and remove it! None of these things cause a decoder to fail any of these tests.
- The DOS command fc /b says there are differences. As mentioned in the Objective decoding test and Least Significant bit accuracy test, the last bit of a decoded mp3 file will vary. What this means is that the result is only accurate to plus-or-minus 1. The full range is 0-65535, so plus-or-minus 1 is pretty insignificant. However, even at that low level, all the recommended decoders sound identical (and correct), but the decodes are not bit-identical. One is not "more accurate" than the other, even though they are different. See the Least Significant bit accuracy test for more details.
- When comparing files using Cool Edit you sometimes get FULL SCALE differences! They're not real though - it's a bug in Cool Edit - see the test methodology for more details.
If you find any other problems, please let me know.
Please see the articles at mp3.com or this introductory guide, this FAQ, or these notes on successful ripping and encoding. They'll be more help than me - I'm not part of the scene ;-).
FhG stands for Fraunhofer-Gesellschaft. Typically, in the mp3 world, FhG refers to The Fraunhofer IIS-A (Institut Intergrierte Schaltungen) who developed the technology that became MPEG-1 layer-3 coding. They are still the primary company in the mp3 encoder/decoder field, and their products are considered the industry standard by many. Though their earlier mp3 encoders cost $199, their fast encoder is available for free within MusicMatch Jukebox, and is found in many other products. They are also heavily involved in the development of MPEG-2 AAC (advanced audio coding) which, like it says, is more advanced that MPEG-1 or -2 layer 3 (ie mp3).
I've read several messages from people who believe that by decoding an mp3 to a wav, and then playing the wav, they will hear a better sound than by playing the mp3 directly. This isn't usually true. Every mp3 player (e.g. Winamp) includes a decoder - it decodes the mp3 in real time before sending the audio data to your sound card. If it has the facility to decode to wav, then it will store exactly the same audio data in a wave file. The same numbers go to your sound card and the file.
If you use one decoder when playing, and another when decoding to wav, then there may be audible differences, especially if one of the decoders has a fault. Some soundcards offer hardware mp3 decoding. Some packages (e.g. Real Jukebox) use a poor decoder when playing mp3 files, and a good one when decoding them to write to CD. However, this is an exception.
Decoding to .wav does not improve sound quality. Most importantly, it does not give you the same .wav that the original mp3 was created from. Mp3 encoding is a lossy process - audio information is lost that can never be put back. Mp3 decoding does not put any data back - it takes a very compressed file format, and converts it into the audio data that the file represents. However, the audio data that was thrown away during the encoding process is not magically returned. The decoded wav file is much larger than the mp3, but it holds the same information. The original wav file contained much more audio information.
If you do wish to improve the sound quality of an mp3 file, it may be possible to process the decoded .wav with a sound editor (e.g. Cool Edit) to make it sound better, but it's just as easy to make it sound worse! I find buying CDs is a good thing.
If you're decoding mp3 files to burn to a CD, you should probably switch to a decoder that doesn't make mistakes (though Music Match Jukebox only failed by the smallest margin). The pure decoders that failed aren't worth using.
I realise there are factors which may influence your choice of player more than sound quality (eg ease of use, features, stability, skins!) so perhaps you shouldn't change. It's worth upgrading from Windows Media Player 6.4 to 7 beta (though this doesn't run on Windows 95 - that's very poor of Microsoft!). If you use Winamp v2.65 or below, consider upgrading to the latest version. If you're choosing a new player, consider those recommended in the Conclusion.
The CD writing packages that decode mp3s on-the-fly make the job of recording mp3s to audio CDs much easier. Of the ones that failed our tests, some of the faults were relatively minor, and you might choose to put up with these rather than go to the trouble of decoding all the files before burning them to CD. Under many circumstances I would be happy with the performance of Easy CD Creator 4 Deluxe. See How to record mp3s onto CD for further details.
Winamp is the most popular mp3 player, but the decoding engine was seriously flawed. Most versions of Winamp before 2.666 used an in-house decoder called "Nitrane". It's notoriously bad (see here and here), as confirmed by our test results.
Winamp has finally been fixed, and our tests of version 2.7 showed that it now decodes mp3s perfectly.
Yes.
- mp3towave was new when I first tested it, and the author happily ironed out the bugs I found.
- Xaudio's author contacted me, requesting a copy of the one lame encoded file that caused his decoder problems, and may fix the bug in a future release.
- Windows Media Player 7 BETA has a problem with the last bit. I was amazed when Microsoft emailed me to find out what it was. As it seems to be the DirectSound processing, rather than a bug in WMP, I think they're happy as it is. I couldn't believe Microsoft were emailing me about a 1-bit error though!
- Winamp used a notoriously bad decoder. Many people knew this well before my tests, but shortly after I revealed just how bad the Nitrane decoder was, Nullsoft began coding its replacement. I think this was as a result of the increased number of complaints Nullsoft received (publicly on their own message boards) after my tests.
David@Robinson.org - but please check through this site first (see contents) - the answer is probably here!
Go back to the Introduction or go on to the conclusion.