[vorbis-dev] Spectral band replication

Dirk Knop dknop at gwdg.de
Thu Jun 14 04:36:30 PDT 2001



Ahoy Stef & List,

there was a technical discussion about SBR in the c't some weeks ago...
well, i serached all c't s from beginning of the year - and it's (of 
course, murphy is always right) c't 11/01, page 16.
as I've got no scanner i can't make that article available (and i don't 
have the time to type it in manually..)

but the essential is, that SBR works following way:

at 64kbps the signal is capped (lowpass) at 10khz. SBR analyses the "base 
frequencies" (whatever that's supposed to mean) and reconstructs the higher 
harmonics (as you tried to) - at mp3pro limited to 15khz. to make this more 
realistic there are some additional information embedded into the mp3 
stream, that cover the "huell"-curve (don't know the translation of that) 
for information about how loud the signal was. the article says nothing 
about patents, but Lars (Stockis) Liljeryd developed that technique for CTS 
(Coding Technologies Sweden). The bitstream is btw. downward-compatible and 
can be played back with normal mp3-players as well (but without SBR of course).

Hope this helps,

Regards,
Dirk aka Koepi

At 13:31 14.06.2001, you wrote:
>Hello!
>I have the following idea about this "spectral band replication":
>
>Harmonic frequencies are integer multiples of the fundamental frequency. That
>means if you take all frequencies in a particular block (say up to 11 KHz for
>a medium bitrate encoding), double them and add them to the original at a
>lower volume, then this results in frequencies up to 22 KHz.
>
>I suppose that taking anything else but harmonics will rather end up in more
>noise. And don't know if that's it what mp3pro does, because I didn't find
>any details, and they probably won't give them out.
>
>I already tried this with music that I downsampled to 11 KHz, i.e.
>frequencies up to 5.5 KHz, and it did sound much better when adding the
>harmonics x2, x3, x4 or even more.

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.



More information about the Vorbis-dev mailing list