Re: Spectrum Estimate functions

[E. Joshua Rigler]

These spectrum analysis functions are great!  I can't believe I never
noticed them before.  Oh well, I guess I learned more by writing my own
similar functions before finding these.  

I have a question though.  What exactly is the difference between the
pyulear and pburg functions?  I get power spectrum estimates that are
similar in shape and scale (using the same inputs, obviously), but
offset by nearly 40 dB!  Try this:


[b,a] = cheby1(4,3,[0.2, 0.4]);
pyulear(filter(b,a,randn(2^12,1)),2,'db');
hold
pburg(filter(b,a,randn(2^12,1)),2,'db');


Are these results to be expected?  I'm afraid I don't really know what
the difference is between the Yule-Walker and Burg methods of
determining an AR model.  When I use my own techniques for determining
the AR model (either based on correlation functions, or a
straight-forward linear regression), and plot their frequency response
with __power, I get results that are actually ~half-way between the
two!  (...and yes, I ran these tests with the same input time series,
derived from the coefficients set in the example above).

These offsets just seem a little suspicious.  I also tried this with
p=3, and got similar results, although the peak response from pyulear
seemed to shift very slightly to the right.

-EJR



On Tue, 2003-05-06 at 18:37, Paul Kienzle wrote:
>
> See pwelch.m from octave-forge:
> 
> http://octave.sourceforge.net/index/signal.html#Powerspectrumanalysis
> 
> Paul Kienzle
> address@hidden
> 




-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------