>
> > > I agree in that encoding in CBR opposed to VBR gives me a feeling of being
> > > more safe :)
> > I wonder where the myth that joint stereo would somehow adversely affect
> > quality comes from. Was it the web pages from the BladeEnc guy?
>
> Actually, it was this mailinglist. I think Mark Taylor (correct me if I'm dead
> wrong!) said something like:
> "at high bitrates, joint stereo can do more evil than good, because the algorithm
> [of detecting similarities in the left/right channels] is not perfect".
>
we all know this, but just to clarify: jstereo means the encoder
is free to choose between regular (l/r) stereo or mid/side stereo
for each frame.
I think mid/side stereo will always improve quality when the left and
right channels are similar. In that case, the mid channel is more
important than the side channel and allocating more bits to the mid
channel will effectively increase the bandwidth. But there are two
problems I've seen with mid/side stereo:
If the L and R channels are very different (and thus there is a lot
of information in the side channel), encoding in mid/side
stereo will creates problems, since noise in either mid or side
channel will be spread to both the L and R channels. Check out
mstest.wav on the lame web page for an example. If this is coded with
mid/side stereo, you will get some very noticeable artifacts. So lame
only tries to use mid/side stereo when the L and R channels are very
similar.
The second problem I've seen is too much switching:
if lame switches back and forth between mid/side and regular L/R,
the switching itself can create artifacts.
The question is, when do the benefits of mid/side stereo
outway the negatives? Until someone figures out the answer
to that question, we'll just take what the FhG encoder does
as the gospel truth: use mid/side stereo only for 128kbs and less.
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
