FPGA Central

Hi,
Have you tried finding the roots of the Z-transform of an autocorrelatio
and noticed the following thing? I take the full lag of th
autocorrelation, calculate the roots of the corresponding Z-transforme
polynomial, and I see that the roots are scattered all over the uni
circle in the Z-plane. Now, I take the same autocorrelation but
truncated one (truncated to some maximum lag) and calculate the roots o
the corresponding Z-transformed polynomial, I see that most of the root
are getting plotted right on the unit circle itself (i.e. most of th
roots have magnitude = 1).
My question is, why does truncating an autocorrelation gets the roots o
the corresponding Z-transformed polynomial exactly positioned on the uni
circle in the Z-plane? The same autocorrelation without truncation doe
not show any such behavior.
I am using MATLAB "roots" function to calculate the roots.
Your explanation will be highly appreciated.
Cheers
Somanath



This message was sent using the Comp.DSP web interface o
www.DSPRelated.com