>
> I made a short clip out of it and you can find it (no permanent link)
>
> http://www.r3mix.net/kingdom.zip 1.330.218B
>
> It's a seven second clip from manowar - kings of metal - kingdom come.
> Don't mind the genre, but it's really sensitive encoding material
> imho.
>
> LAME382/Dmitry -V1 -mj -h -b128 gives:
> ----- bitrate statistics -----
> [kbps] frames
> 128 1 (0.3%)
> 160 76 (24.8%)
> 192 191 (62.4%)
> 224 25 (8.2%)
> 256 13 (4.2%)
> 320 0 (0.0%)
> average: 189 kbs
>
> I believe this is a severe case of:
>
> chirps: high-frequency distortions added to the audio, will sound like short chirps,
>whistles, sometimes can shwoosh up like a high-pitched glissando (slide).
>
> as David McIntyre would say. I would say: the sharp high-squeeling noise
> that makes the back of your skull/neck hurt and just about the single
> most painful artifact on the block.
>
> Listen to the "K" from "Kingdom" and the (cymbals?).
>
Does the chirp sound like an encoding glitch, or could it be explained
by a lack of bits? At these high bitrates, there are two possible
bugs you could be running into: There was a joint stereo bug
introduced in 3.81 and fixed in 3.82 beta. If the version you are
usings says "3.82 alpha", it means it comes direct from CVS and it
might or might not have this bug.
The other bug might be this relaxed not-quite-ISO-spec
framesizes that lame 3.82 can generate. So if you dont mind,
can you try the following: (I would try this my self, but
I'm guessing these chirps are too subtle for me to detect :-)
CBR @ 128kbs. If this also has the problem, it is just the
fact that these cymbals need more bits. It would also be
interesting to report on how mp3enc at 128kbs compares.
For preecho type stuff like this, mp3enc is usually a little
better than lame.
LAME382/ -V1 -mj -h -b128 --strictly-enforce-ISO
If this fixes the problem, then we need to go back to enforcing
the exact ISO buffer requirements.
LAME382/ -V1 -m s -h -b128
If this fixes the problem, it could be the jstereo bug.
> S/JS makes no difference @256 and I don't really hear a difference, or
> let's say I can enjoy the clip 100% at 256kbit/s. It might have a
> sharper feeling to it, but nothing to be disturbed about.
>
Since JS @256 seems to be working, it probably it not the JS bug.
>
> btw: I also did a quick meaningless freq analysis, and the 256 is
> right up there until 20kHz, and the VBR is off the scale since exactly
> 16kHz. What defines this magical 16kHz? (or is this just a big
> coincidence?)
>
the last critical band is 16-22khz, and this band does not
have a scalefactor, (and psycho acoustics are not computed
for it either) so it is hard to control the noise in this band.
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
