polyphase filter bank : phase shift and aliashing reduction Options

[hkato111]

Hello,

this is a Japanese to try to summarize the mathematics for MP3.
If you know the answer to the below question, could you reply?

question:
Why the phase shift of polyphase filterbank of MP3 could corrected by multipled -1
to the output of the filterbank when k is odd and v is odd, where k is subband index
(k=0,1,2,..,31) and v is subband time index (real time t = 32*v) ?

This phase shift was noted in
B. Edler, Aliasing reduction in sub-bands of cascaded filterbanks withdecimation, Electronic Letters, vol. 28, no. 12, pp. 1104-1106, Jun. 1992
and this question happened to check (3.5) upper equation of the paper
http://www.mp3-tech.org/programmer/docs/paper-1-english.pdf
on http://www.hydrogenaudio.org/forums/lofive...php/t53319.html.
(Thank you , Garf-san)

I can only imagine that multiplied -1 to the output of the filterbank when k is odd
and v is odd means multiplied (-1)^(kv) to the equation of polyphase filter bank.

Best Regards,

e-mail

[SebastianG]

question:
Why the phase shift of polyphase filterbank of MP3 could corrected by multipled -1
to the output of the filterbank when k is odd and v is odd, where k is subband index
(k=0,1,2,..,31) and v is subband time index (real time t = 32*v) ?

I wouldn't say "phase shift". It just flips the spectrum. Low frequencies become high frequencies and the other way around. It's done to compensate for the fact that downsampling for every odd band intrinsically flips the spectrum upside down.

Cheers!
SG

[hkato111]

Thank you for your reply and kindness.

>It's done to compensate for the fact that downsampling for every odd band
>intrinsically flips the spectrum upside down.

Would you show the web page to explain that if you know?

I cannot find the web page like that by myself...


H.Kato