Howdy Gabriel,
> Am I the only one wich doesn't understand what Frank is telling?
> Can anyone explain a little?
You are not alone. Once I get past his aggressively authoritative posting
style (which is a major disincentive to understanding for me), I find I
understand about half of what Frank posts. (Some of which seems actually
both relevant _and_ useful.*) This is not one of those postings.
I assume he has some theoretical basis for wanting to correlate
differentiated signals (rather than raw signals), and that this is somehow
relevant to LAME (I haven't grepped through the source, but I remember
hearing some discussion of a correlator being used for joint switching?),
but, since he mentions neither before reeling off a long tabulation of
statistics, it all seems pretty out of the blue to me, too.
Perhaps a better question would be: Does anyone actually understand what
Frank is _saying_ (not tell), and would they care to enlighten the rest of
us?
Alex
*A lot of Frank's commentary seems to point to (possibly) radical areas for
improvement. (Case in point - that rant about dithering, which would be
fairly difficult to incorporate [as dither noise shaping would have to be
integrated into the encoder distortion calculations as well as the decoder],
but potentially very useful. Also, the bit about probabilistically
measuring the 'distortion of the undistorted bands' - I want to check out
that paper.)
But a similar amount seems clueless, useless, or counterproductive. (For
instance, single-handedly taking on the task of making LAME C++ compatible,
when no one else has expressed any desire for it, and when it means massive
diffs against the original - no coder with any experience working on a major
project _with other people_ would ever do this lightly.)
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
