Re: Bandpass system identification using IIRs (audio)

On Mar 19, 3:49 am, "steph_tsf" <stephanecnocka...@xxxxxxxxxxx> wrote:
I'd like to identify how and why FDLS_200 is turning away from a gentle
highpass, knowing the results are outstanding with many other kind of
filters, even complicated filters with a linear combination of LPF, BPF
and LPF outputs numerators) as showed previously in this thread.

I'm not sure I would call it a "gentle" highpass, when the
poles and zeros are so close to each other and to the unit
circle. The relative passband of the HPF is very different
from your LPF experiment.

FDLS is a least squares interpolation method, subject
to various types of interpolation and numerical errors.
You've specified the response relatively very far away from the
location of the poles and zeros, and with only 6 digit precision,
and are trying to interpolate a frequency response which looks
like it requires many more bits of precision to specify the
spacing of the zeros from each other and the poles. The exact
location of zeros looks like it is even more sensitive to
errors in the specification of distant frequency response
points, than the poles. I would suggest specifying the
phase/frequency response locations much closer to the location
of the zeros (and the poles), and with much higher precision.

Perhaps an iterative method of using FDLS would work, where
you initially estimate the location of the poles and zeros,
and dynamically adjust the location of the response points,
or just increase the number of them, until there is less
sensitivity to interpolation errors from noise in the
sample point specifications.




IMHO. YMMV.
--
rhn A.T nicholson d.0.t C-o-M
.

Relevant Pages

  • Re: Iterative Technique for PID Controller Tuning
    ... Page 5 The point to point response for my 3 poles at -2*PI*10 is plotted. ... The response is a critically damped response except for when the output is saturate. ... Notice that poles are were I put them but as I said before, the zeros can go anywhere. ... Page 8 The closed loop Bode plot. ...
    (sci.engr.control)
  • Re: Iterative Technique for PID Controller Tuning
    ... Page 5 The point to point response for my 3 poles at -2*PI*10 is plotted. ... The response is a critically damped response except for when the output is saturate. ... Notice that poles are were I put them but as I said before, the zeros can go anywhere. ... Page 8 The closed loop Bode plot. ...
    (sci.engr.control)
  • Re: IIR and FIR filters
    ... This was wrong - too quick of a response. ... An IIR has non-trivial poles (it could have non-trivial zeros as well). ...
    (comp.dsp)
  • allpass filter
    ... response is known, but not the poles or zeros. ... Can the function 'invfreqz' be used or is there some other ...
    (comp.soft-sys.matlab)
  • Re: s-plane; Laplace; Smith book
    ... using my old standard circuit book. ... Fig 30-9 again has the poles on the LS. ... where both the probe and impulse response decrease. ... I am quite familiar with sigma + j-omega plane as tool ...
    (comp.dsp)
Loading