Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!usc!samsung!think!mintaka!mit-eddie!uw-beaver!ubc-cs!alberta!calgary!!smit
From: smit@.ucalgary.ca (Theo Smit)
Newsgroups: comp.dsp
Subject: Re: AR and MA questions
Summary: Excuuuuuuuse me!
Message-ID: <2172@cs-spool.calgary.UUCP>
Date: 27 Nov 89 16:15:25 GMT
References: <1389@mrsvr.UUCP> <2074@cs-spool.calgary.UUCP> <3556@quanta.eng.ohio-state.edu>
Sender: news@calgary.UUCP
Reply-To: smit@enel.ucalgary.ca (Theo Smit)
Organization: U. of Calgary, Calgary, Alberta, Canada
Lines: 26

In article <3556@quanta.eng.ohio-state.edu> rob@kaa.eng.ohio-state.edu (Rob Carriere) writes:
>In article <2074@cs-spool.calgary.UUCP> smit@enel.ucalgary.ca (Theo Smit)
>>Yes. 
>
>Wrong.  The AR parameters are the coefficients of a polynomial whose _roots_
>are the poles.
[...]
>>Right again. [...]
>
>Wrong again.  The zeroes are the _roots_ of the polynomial described by the
>MA parameters.
> SR

OK, so I goofed. Anyway, what Rob says is what I _meant_. Most of the time
the AR and MA polynomials are more useful than the explicit pole locations
(i.e. doing an FFT on the coeffs to get the periodogram of the model), so
I usually don't bother finding the actual poles.

On a different note, doing FFT's on the AR or MA parameters is computationally
much less intensive if you use the interpolating FFT I described earlier.
(that would be the Screenivas-Rao algorithm, not mine; mine works well for
interpolating time sequences when you start out in the frequency domain).
For all of you who responded to my previous posting on the EZFFT, I will be
posting the source and some more explanation shortly. Stay tuned.

Theo Smit (smit@enel.ucalgary.ca)


Brought to you by Super Global Mega Corp .com