[vorbis-dev] Book descriptions?

[Michael Smith]

At 10:47 AM 8/28/00 -0600, you wrote:
>Can anybody tell me where I can find some info on the new books? I noticed
>there are now A-E. What exactly does this mean? How does it translate to
>higher and lower sampling frequencies and bit rates? Ideally I'd like
>something in the 11kHZ, 8bit mono area...
>
>I also noticed this struct:
>/* CD quality stereo, no channel coupling */
>vorbis_info info_B={
>  /* channels, sample rate, upperkbps, nominalkbps, lowerkbps */
>  0, 2, 44100, 0,0,0,
>  /* smallblock, largeblock */
>  {256, 2048}, .....
>
>Does this mean I can just start tweaking away by changing the channels,
>sample rate upper, nominal and lower? I'm guessing no, since I'm pretty sure
>this must screw up some math somewhere...how does one actually tell the
>encoder that they want these settings?

Yes, you can change any of those settings. However...
        sample rate: changing this will work correctly, but the codebooks aren't
optimised for other sample rates, so the quality will be slightly lower
than it would be if you built new codebooks (building codebooks is rather
complex, and the tools don't run on windows - you probably don't want to do
this). Play around with it - if the quality loss is acceptable (and it'll
be fairly small, probably), then go for it. For lower sample rates, I think
there will be appropriate books once vorbis has better low-bitrate support.
        channels: ok. OggEnc changes this to deal with mono files. More than 2
channels is also possible, but don't expect players to deal with such files
correctly.
        upperkbps/nominalkbps/lowerkbps: change these all you want. Currently,
they're ignored. That's one of the major features needed before vorbis hits
1.0

As for what the different modes (A-E) are - they're just sets of
psychoacoustic settings and codebooks for a set of different bitrates.
Typically, A->128, B->160, C->192, D->256, E->350 (kbps, for stereo).

I think Monty intends to add a better way of selecting bitrates
(programatically, rather than just picking a single fixed mode) at a later
date, but for now, just pick one of these and use it. See oggenc for
example code (which is somewhat messier than it needs to be, as a result of
being written at about 3am the night before beta2, but which works correctly).

Michael

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/