Many MP3 Codecs are floating around the Web. Xing is a big one, as is LameEnc. Many freeware MP3 ripper/encoder combos use the Xing codec (Audiocatalyst, MpegDJ, and Music Match to name a few). This presents a problem to those of us who give a damn about the quality of MP3 files. Simply put, Xing sucks. It crops frequencies thereby destroying quality. Herein data will be presented illustrating this point.

Xing also destroys lower-level stereo panning information which also serves to lower the quality of MP3 files. Although I will not present any data to this end - take my word for it! If you are still interested, please click here for someone else's tests

Cool Edit Pro 1.2 was employed to generate a test file consisting of a frequency sweep. The file was saved in wav format, and then encoded to with LameEnc, and the Xing codec (included in Audiocatalyst 2.1). Each of the mp3 files are encoded to 192 kbps, HQ, 44,100Hz using Stereo mode.

The specific parameters of the test file are as follows:

A starting frequency of 10 Hz was ramped up to 26,000 Hz (26 kHz) over a 20 second time period. This tone sweep was generated at a sampling rate of 44.1 kHz, with a resolution of 16-bit and sampled in full stereo (just like a CD).

A Fourier Transform (FT) was applied to each of files (the original wav, the mp3 encoded with LameEnc, and the mp3 encoded with Xing.) The Fourier transform converts a signal value that is a function of time, space, or both into a function of frequency. In laymen's terms, it translates something time dependent (like sound), into its respective frequencies. For example, if you play a "Middle A" on a piano and perform a FT on the sound (assuming your piano is in tune), you should see a sharp spike at 440 Hz (the frequency for Middle A). In our case, we should see signals from 10 all the way to 26 kHz for each file. As you will see, this is not the case.

Shown in purple above, is the original wav form. Notice it is more or less constant from the low end all the way up to 26 kHz. This indicates that the entire file contains all the frequencies. Shown in a light periwinkle is the same data encoded w/ LameEnc. Notice signals all the way to the to 26 kHz. This demonstrates preservation of frequency and quality.

Shown in teal above is the same data encoded with the Xing codec. Notice the sharp cut off at 16 kHz! Most people can hear sounds above 16 kHz. Xing distorts and cuts-off anything with a frequency above this value. It's no wonder high frequency sounds (such as hi-hats and cymbals) sound messed-up when encoded with this inferior quality software. This is, in part, why some MP3 files sound "swishy." Bitrate is also a variable in this effect, as is stereo mode, but the cut-off phenomenon is ubiquitous to all bitrates, and is as well independent of stereo mode.

Conclusion

Xing sucks. It distorts and cuts-off frequencies >16 kHz. This is NOT CD quality by any stretch of the imagination. LameEnc on the other hand preserves the entire frequency range of the music file. (Also, it does the best job at presenting stereo panning information, although I did not present any data to this effect.)

Again, there is MUCH more to this than the simple data I showed above. If you are REALLY interested in this, I would strongly recommend a book entitled, MP3: The Definitive Guide by Scott Hacker. Several chapters of this book are actually available on-line

Well, now you know better! If you know someone who is encoding with shit software, please, direct them to this page and instruct them to READ IT and DO IT!