Any shape IIR Filter

"Greg Berchin" <gberchin@comicast.net> wrote in message
news:1k15i3l9ccr1mtsa84i60apo2e32fvlt82@4ax.com...
>
> FDLS requires phase information in addition to magnitude; not always
> easy to come by.  You can derive the minimum phase characteristic from
> the magnitude, however, and FDLS should work fine with that.

Greg,

Is the result of FLDS always a stable function?

VLV

On Sun, 28 Oct 2007 07:35:25 -0600, "Vladimir Vassilevsky"
<antispam_bogus@hotmail.com> wrote:

>Is the result of FLDS always a stable function?

In general, the model has the same stability as the prototype.
Exceptions occur when a pole pair is very near the unit circle, in which
case the least squares nature of the fit can push them "over the edge".
It's rare, but it does happen.

Also, and this part got cut out of the magazine article, if you use sine
waves instead of cosine waves to fill the matrices, you're almost
guaranteed to get a pole at or near z=-1 (often outside the unit
circle).  This happens because sampling a sine wave of frequency Fs/2
only catches the zero-crossings.  The interpretation is a
"zero-input/nonzero-output" situation, and the mathematics therefore
place a pole at half the sampling frequency.  The cure for this is to
use only cosines in the implementation.

(It can be shown that the cosine formulation models only the real part
of the frequency response, and that the sine formulation models only the
imaginary part.  If you use only the cosine formulation, you can gain
access to the imaginary part by artificially introducing some delay into
the system.)

Greg

Hi Jeff,
Here is one way: use a neural net learning algorithm to match the desired
response.  See page 476 of Chapter 26 in my free on-line DSP book:
www.DSPguide.com.    
Regards,
Steve Smith

SteveSmith wrote:
> Hi Jeff,
> Here is one way: use a neural net learning algorithm to match the desired
> response.  See page 476 of Chapter 26 in my free on-line DSP book:
> www.DSPguide.com.  

Thanks for that on-line book. It's been a great help to many.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

On Nov 1, 8:28 am, "SteveSmith" <Steve.Smi...@SpectrumSDI.com> wrote:
> Here is one way: use a neural net learning algorithm to match the desired
> response.  See page 476 of Chapter 26 in my free on-line DSP book:www.DSPguide.com.

There are also genetic optimization algorithms
that have been applied to IIR and FIR filter design,
such as "differential evolution".



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

Ron N. wrote:
> On Nov 1, 8:28 am, "SteveSmith" <Steve.Smi...@SpectrumSDI.com> wrote:
> 
>>Here is one way: use a neural net learning algorithm to match the desired
>>response.  See page 476 of Chapter 26 in my free on-line DSP book:www.DSPguide.com.
> 
> 
> There are also genetic optimization algorithms
> that have been applied to IIR and FIR filter design,
> such as "differential evolution".
> 

How many pompous buzzwords were invented to describe a dumb search by 
brute force :)

VLV

On Nov 1, 8:00 pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com>
wrote:
> Ron N. wrote:
> > On Nov 1, 8:28 am, "SteveSmith" <Steve.Smi...@SpectrumSDI.com> wrote:
>
> >>Here is one way: use a neural net learning algorithm to match the desired
> >>response.  See page 476 of Chapter 26 in my free on-line DSP book:www.DSPguide.com.
>
> > There are also genetic optimization algorithms
> > that have been applied to IIR and FIR filter design,
> > such as "differential evolution".
>
> How many pompous buzzwords were invented to describe a dumb search by
> brute force :)

well, pompous or not, i like genital optimization algorithms.

--

Emily Litella

robert bristow-johnson wrote:
> On Nov 1, 8:00 pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com>
> wrote:
>> Ron N. wrote:
>>> On Nov 1, 8:28 am, "SteveSmith" <Steve.Smi...@SpectrumSDI.com> wrote:
>>>> Here is one way: use a neural net learning algorithm to match the desired
>>>> response.  See page 476 of Chapter 26 in my free on-line DSP book:www.DSPguide.com.
>>> There are also genetic optimization algorithms
>>> that have been applied to IIR and FIR filter design,
>>> such as "differential evolution".
>> How many pompous buzzwords were invented to describe a dumb search by
>> brute force :)
> 
> well, pompous or not, i like genital optimization algorithms.

I get at least a hundred spam e-mails a day offering that. Would you 
like me to forward them to you?

Steve

Steve Underwood wrote:

>>>> There are also genetic optimization algorithms
>>>> that have been applied to IIR and FIR filter design,
>>>> such as "differential evolution".
>>>
>>> How many pompous buzzwords were invented to describe a dumb search by
>>> brute force :)
>>
>>
>> well, pompous or not, i like genital optimization algorithms.
> 
> 
> I get at least a hundred spam e-mails a day offering that. Would you 
> like me to forward them to you? 

You might want to look at this: www.spampal.org

Seriously speaking, the name of the Greg Berchin should be placed in the 
pantheon somewhere in between Parks and McClellan :)

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

On Nov 2, 11:27 am, Vladimir Vassilevsky <antispam_bo...@hotmail.com>
wrote:

> Seriously speaking, the name of the Greg Berchin should be placed in the
> pantheon somewhere in between Parks and McClellan :)

Whoa; I'm seriously flattered, but I took a class from Tom Parks a
long, long time ago and I don't think that I could even tie his shoes,
metaphorically speaking.

I take it, then, that you've had success with the algorithm?

There is so much about that algorithm that never got published.
Problem is, it's been so long since I studied it that I don't remember
most of it.

Greg