MP3 Quality

[Brahmb]

I've just downloaded some MP3s from Napster (only ones that I already have the CD's of- I'm not sure where I fall on the copyright issue yet). I'm listening and comparing the quality of mix. My questions are:
Is there a difference in quality from MP3 to MP3? I know from CD to CD varies, and as I listen to early Billy Joel vs. mid-career on MP3 I notice a hiss on the earlier. Is this CD quality of the person who originally uploaded the MP3? Is it the quality (or lack there of) in the earlier recording? I notice a difference on the CD's but not as markedly. Any help would help me with my mixing decisions.... Thanks.



[This message has been edited by Brahmb (edited 07-29-2000).]

[drstawl]

Besides the variability in recording quality from CD to CD that you mentioned, several other factors determine the quality of the resultant .mp3's.
When CD files are converted to .wav files only a digital transfer will give you most of what was on the CD. Some people doing this go through two more conversions (D/A and A/D) to get it to .wav format. Some use a mic in front of their speakers! Some use crappy "ripper" programs that introduce some noise.
Then you have the errors produced by various encoders. Add to this the unavoidable losses at various .mp3 bitrates and you see the number of ways to screw it up are large.

[John Sayers]

I've just read an article on Mp3 encoders - it's worth a read
http://www.arstechnica.com/wankerdes...mp3/mp3-1.html

[pglewis]

Brahmb: Here's the skinny. Every encoder is different. One encoder might sound better than another to you on one song, but the reverse might be true on a different song. MP3 uses perceptual techiniques. There is a lot of information in music that your ears never really detect... especially during the loud parts (you know, amp hum always gets buried). This is one of the things MP3 takes advantage of to save space. It literally throws away some of the music, and the original cannot be exactly reconstructed from the resulting MP3. It is called lossy compression because there's no way to exactly recover what was thrown away. The "bitrate" is the biggest factor for quality with any encoder. At lower bitrates, you have smaller files... but throwing away more information generally results in "artifacts" (those metallic ringing or underwater sounds you've undoubtedly heard in MP3's). As the bitrate increases, so does the quality... but the resulting file is larger. You trade quality for size.

John: I've seen that one, actually. FWIW, their tests were with a somewhat old version of LAME, and I don't think they specified the parameters used to encode in all their tests. My personal listening tests support the information I recently found at: http://www.r3mix.net/ -- you might be interested in that one. If you go to the "analysis" section there, they have a few valid reproducable tests. But you don't judge art with your ears, so I don't trust too many graphs for my music. Using LAME with the VBR settings suggested there, I'd have a tough time distinguishing the MP3's created vs. my wave source. I haven't done a true double-blind test, but I'm usually pretty picky. Since the space required is usually just a little more than a 192kbps file (YMMV), and I've yet to hear encoding artifacts... there is nothing better short of lossless compression right now, IMO.