Hi

In Roberto's June 2004 32 kbps (dial-up bitrate) multi-codec listening test, Ogg Vorbis post-1.0.1CVS --managed -b 32 resampled with SSRC was used. Unfortunately, it didn't perform too well, probably due to its default downsampling to 22 kHz and the consequent, rather stringent lowpass, whereas the competition (at least in this particular listening test) was applying a 32 kHz sampling rate.

Which leaves me with the question: which quality setting (be it CBR or VBR) would yield better results at this target bitrate? I would be interested in hosting a live audio stream on a radio station's website in Vorbis at a 32 kbps-like bitrate. I haven't quite decided for mono or stereo yet, so I'd like to test it out for myself.

I'm really targeting dial-up modems, which is why a potential VBR encoding shouldn't fluctuate too much above 32 kbps to prevent bottlenecks. The next bitrate step to 32 kbps being 40 kbps (i.e. too high for this purpose), I guess 32 kbps CBR would perhaps be my best option? Or would it make any sense to set encoding to -M32 (with or without --downmix to mono)?

So any recommendations for this 32 kbps Vorbis setting, both mono and stereo? I hardly have any experience at all with Vorbis, let alone at rates this low, so thanks in advance for your advice.

Fist of all, get the latest aoTuV encoder. Stay away from libvorbis 1.0.1 and 1.1 (both have no great 32 kbps mode).

Try -q -2 with 44.1 kHz in stereo. If this doesn´t suite your quality needs you can try mono input and a higher -q value. I don´t think the 32 kHz tunings are as nice as the 44.1 kHz ones - so I don´t expect better performance with resampling. That´s actually a wild gues - so you might try this, too

I didn´t observe big bitrate fluctuations in -q -2 mode that may make the client´s buffer running too low. Only enforce ABR if absolutely necesssary.

We're talking about this right over here: http://www.hydrogenaudio.org/forums/index....opic=31807&st=0 =)