[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: AUDITORY Digest - 30 Oct 2003 to 6 Nov 2003 (#2003-183)

Dear Hafiz
An all-pass filter is only flat if the entire impulse response is
processed. When you take a finite window of a filtered stimulus,
this rule is broken. In a sense, this is equivalent to taking a piece of
the impulse response a(t) and expecting it to have a constant spectrum.
Such a piece of a(t) will not be flat. The longer you make a(t), the
worse the problem becomes.
Hope this helps,


Automatic digest processor wrote:

There is one message totalling 46 lines in this issue.

Topics of the day:

  1. why all-pass filter introducing inaudiable magnitude distortion


Date:    Thu, 6 Nov 2003 03:20:49 -0600
From:    hmalik1 <hmalik1@UIC.EDU>
Subject: why all-pass filter introducing inaudiable magnitude distortion

Dear Experts and Friends,
I need guidance for understanding the theory behind distortion(inaudible)
introduced by all-pass filter.....

I'm using 10-14th order all-pass filter(cascade form) with complex pole-zero
pair close to unit circle i.e.( pole at .9 and zero at 1/.9 and theta = +/-
when a frame of 128-256 audio samples passed through the all-pass filter and
power-spectrum of the output frame is estimated using
a) Moving Average signal model
b) Autoregressive signal model

this is observed that all-pass filter has introduced a local minima in the
estimated power spectrum at frequency  = phi for both models, but minima in
model (a) is stronger than model (b).

this distortion introduced by all-pass filter become more stronger as
pole-zero pair moves close to unit circle or by increasing the order of the
filter used. (simulation is done using MatLab)

I would appreciate any suggestion or reference that'll help me to understand.
this behavior .

thanks in advance.


Hafiz M.A. Malik,
Multimedia Systems LAB.
Elecrical & Computer Engineering Dept,
UIC, Chicago, IL, 60607


End of AUDITORY Digest - 30 Oct 2003 to 6 Nov 2003 (#2003-183)