[vorbis-dev] Thought for the new year

Segher Boessenkool segher at wanadoo.nl
Wed Dec 27 16:11:00 PST 2000

Frank Klemm wrote:
> ::  Gregory Maxwell wrote:
> ::  >
> ::  > On Wed, Dec 27, 2000 at 04:41:55AM +0100, Segher Boessenkool wrote:
> ::  > > 1) MDCT is good for image coding
> ::  > > 2) image coding and audio coding are two very different things
> ::  > > 3) combine 1 and 2
> ::  >
> ::  > It follows that MDCT is good for many differnt tasks. No? :)
> ::
> ::  "good" is subjective ;-) Hey, I said for the new year, you're too
> ::  early!
> ::
> A very rough estimation of quality can be get by:

By "good for coding" I don't mean quality, of course; we're talking
about a transform for compression here, so a "good" transform is one
that makes it easier to compress, i.e. it should group the features
of the sound together based on some sensible norm. MDCT uses a window
that doesn't change size for differing frequencies; that's no good for
a perceptual codec.

>                             used_bitrate
> sounds_like_bitrate = -------------------------
>                       sqrt ( number_of_codings)
> while using the lowpass frequency of the used_bitrate, not of
> the sounds_like_bitrate.

Man, you're using a lousy encoder! It should stabilize...

> ::  Even if unstable: "almost" the same results.
> ::
> A 100th generation MP3 encoding sounds like
> laf"& f&f(tr dq(&/u)ipw(& d-.jwe(;qr d:_&jqrj_:I&oq qej

That must have been difficult to type. What language is it?  ;-)

> Losses become less for every additional generation, you're right.

That's not what I was saying.

> Because there's nothing which can be lost. You can destroy trash.

> This doesn't help. You got the errors of the adjacent granule. This is

You can compensate for that (see MDCT_CORR in my latest Vorbis patch).

> enough so that on average 40% of the spectral coefficients are different
> from the previous pass. You are smearing the signal into the time and
> frequency domain. This is enough so you get different results in every pass.

So you _do_ agree that MDCT is a lousy transform for audio coding?



--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.

More information about the Vorbis-dev mailing list