Tamij wrote:
> Dear All,
>
> I am DSP engineer working on MP3 decoder implementation. I got
> doubts while analyzing subband synthesis and IMDCT. In between these two
> modules, frequency inversion is done for IMDCT output samples. Why is it
> done here??. What will happen if it is not done? I have searched many
> forums but couldn't able to get. Eagerly expecting your response.
>
The point here is that the alias reduction step, that combines two
freuencies of adjacent subbands (the original and the suspected alias),
will work only, if the phase difference is correct. To achive the correct
phase, the frequency inversion step is applied (and its effect must be reversed
in
the decoder). For a somewhat more detailed explanation I could recomend
the book "Undestanding MP3" (sorry, I usually do not like self advertisement,
but if
you work on an MP3 decoder, it is probably a good investment anyway) or you can
consult the PhD thesis of Bernd Edler for a very detailed treatment (in German)
or
the article in the Electonic Letters, both cited below.
Yours
Martin Ruckert
@PhdThesis{Edler95,
author = {Bernd Edler},
title = {\"Aquivalenz von {Transformation} und {Teilbandzerlegung} in
d
er {Quellencodierung}},
school = {University of Hannover},
year = {1995}
}
@Article{Edler92,
author = {Edler, Bernd},
title = {Aliasing Reduction in Sub-Bands of Cascaded Filter Banks with
Decimation},
journal = {Electronic Letters},
year = {1992},
volume = {28},
number = {12},
pages = {1104--1105},
month = {June}
}
_______________________________________________
mp3encoder mailing list
[email protected]
http://minnie.tuhs.org/mailman/listinfo/mp3encoder
