>
> After 8-9 hours of trying to grasp a fraction of the source (no
> programmer, no psy-coding background, plain dumb :)) (in a futile attempt to find
>the reason why
> "-b" is misinterpreted (?)) this catched my eye:
>
> qadjust=-2.5; /* start with -1 db quality improvement over quantize.c VBR */
>
> in Mark's vbrquantize.c. Why this "-2.5"?
>
> Anyways, I changed the thing to some other values and
> "qadjust=-.5;" combined with "-q1" gives filesizes/bitrate
> distributions very much like the --old-vbr. (meaning: V1,JS=transparent
> for most, at 170-180kbit/s average)
>
You are right - that is a tunable parameter that just adjusts the -V
scale. There is no particular quality associated with -V1 (or any
other setting) - it is just a sliding scale that can be adjusted to
anything you want.
With VBR (both modes), each frame is encoded with as few bits
as possible so that:
quantization_noise < allowed_masking - N db
and the -V setting (and that quality adjustment) just change
the value of N. (allowed_maskings are the maskings computed
by psymodel.c)
The goal is something like this:
VBR_q compression like
4.4 320kbs/41khz
0 5.0
5.5 256kbs/41khz
1 6.0
2 7.0
7.3 192kbs/41khz
3 8.0
8.8 160kbs/41khz
4 9.0
6 11 128kbs/41khz
9 14.0
14.7 96kbs
So -V6 is supposed to give, on average, 128kbs.
-V0 is supposed to give, on average, 256kbs. This doesn't really
make sense because you might as well just use 256kbs CBR, but this
is because many people expect that -V0 should give the best possible
quality. -V1 should be around 220kbs, and with another 20% reduction
from jstereo, that is down to 180kbs.
All the various default settings (jstereo, lowpass filtering) are
based on the compression ratio. For VBR, the compression ratio is not
known in advance, so LAME uses the above table.
>
> Negative:
> - some nasty artifacts in a file I encoded (every vbr_mt does
> worse than vbr_rh on this file), I'll address this in another
> thread. (*, if desired)
yes, please send me the .wav files with a description of the
problem. I have two samples already with artificts, but I
haven't had a chance to figure out what is wrong.
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
