Log in

View Full Version : slightly tech. question about quality lose


cochese
17th January 2004, 05:20
I have a decent enough understanding of mp3 encoding & the theory on how it works. But from a few posts I have recently, I have begun to get confused about one point.

Say I have a mp3 file and the wav file it came from. I get why the mp3 file has quality lose compared to the wav file it came from. Now, if I take that mp3 file and decode it back to it's wav file, then re-encode the wav back to the mp3 file...will I get quality lose?

From *my* logic, it seems that when I decode the mp3 file back to a wav, that second wav will not be of the same quality as the original wav (will be less). Right? So if I take the wav decoded from the mp3 and then encode it back to the mp3 (at the same bitrate), the quality of the 2nd encoded mp3 should be the same as the mp3 from the first time?

If anyone is able to follow this ( i know, it's a bit confusing ) I would love a response. Thank you.

KpeX
17th January 2004, 05:29
Welcome to the forums!

Theoretically I suppose it is possible, but it does not work that way in practice.

Assuming a WAV source such as from CD, when you first encode this audio to MP3 loss is introduced in the form of artifacts, which may or may not be audible, since MP3 is a lossy encoding process. When you decode the MP3 to a WAV file, that WAV has the same 'quality' as the encoded MP3, and contains the same artifacts. If this wav is then reencoded to MP3, more loss will be introduced and more artifacts produced in the audio, which will be very likely to be different from the first MP3 encoding.

Thinking of it another way: every lossy encoding process attempts to reproduce the signal it is encoding as close as possible to the original , but must introduce artifacts (noise) when necessary according to bitrate. When reencoding MP3, the encoder attempts to reproduce that signal as closely as possible as well, so the quality is lowered once again.

cochese
17th January 2004, 16:11
I follow ya. I think I'm seeing bitrate as a measure of quality, and not as a measure of size (which it is basically).

Mug Funky
19th January 2004, 10:53
wouldn't it be theoretically possible to get the same mp3 back if you decode to 32bit wav, and use the exact same encoder/settings?

note to self: i should try this.

cochese
19th January 2004, 15:55
Note to Mug Funky: i'd like to know :)

KpeX
19th January 2004, 16:51
Originally posted by Mug Funky
wouldn't it be theoretically possible to get the same mp3 back if you decode to 32bit wav, and use the exact same encoder/settings?

note to self: i should try this. I don't think so, because the psymodel should react differently to every audio - and the MP3 decoded WAV should be significantly different than the original wav. Although it's an interesting idea.