[TriLUG] Ogg vs. mp3

[Tanner Lovelace]

On Fri, 2003-05-30 at 01:55, al johson wrote:
> Actually, according to various sound experts, as long as you sample MP3
> files at 160-180 most experts really can't detect the difference between MP3
> files and Wav files.
> While a Wav file clearly has more data, you must remember that there is
> sampling taking place simply in the conversion between analog to digital and
> back again, when even wav files are played. There was a math professor at
> the University of Pittsburgh who used to play a trick on her students to
> teach them about sampling. She would go to the CD store and buy the newest
> and most popular album available. She would take it to her math class, but
> before she would play it she would put some bad scratches on it. The class
> really payed attention to that, but when she played the CD there was no
> trace of the scratches, because she hadn't scratched it enough for it to
> become audible. She then explained to the students why CD's don't act the
> same as 33 records. This was a story that was told to me in the Pitt alumni
> magazine (I did my graduate work there).
>     So as long as you sample at a sufficiently high enough rate there is no
> audible difference between sound when it is sampled or not. Hence the magic
> numbers I've given above.--Al Johnson

Actually, Al, I believe that the reason you can scratch a CD and still
have it play is that the information is stored redundantly on the CD.
Also, the sampling taking place in the analog to digital conversion
isn't all that applicable because that sampling is above the 
Nyquist limit. (The Nyquist limit says that in order to accurately
reconstruct a signal you must sample at a rate of at least twice
that of the signal to be reconstructed.) 

Also, as far as the difference between wave files and mp3, you have
to remember that MP3 is a *lossy* format.  That is, it throws away
frequencies that it thinks that *most* people won't hear.  Notice
I said most people.  As in anything statistical, there will be people
who can hear those artifacts.  In addition, it will vary from song
to song.  I once found one song that I couldn't accurately convert
to MP3 even at 256kbits/sec.  Because of Nyquist, if you've got 
less bits, you won't be able to accurately reconstruct the signal
no matter what you do.

Now, that said, MP3s at 192kbits/sec are adequate for most songs
and people out there.  I don't know that much about ogg vorbis
so I can't say if less bits there is ok.  Perhaps someone who 
understands how vorbis does the compression can mention something
about that?

Cheers,
Tanner
-- 
Tanner Lovelace |  lovelace(at)trilug.org  | http://www.trilug.org/
--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--
GPG Fingerprint = A66C 8660 924F 5F8C 71DA  BDD0 CE09 4F8E DE76 39D4
GPG Key can be found at http://wtl.wayfarer.org/lovelace.gpg.asc
--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--*--
 This would be a very good time to hang out with the Open Source 
 people, before they get formally reclassified as a national security 
 threat. -- Bruce Sterling