I need to solve the following problem: I was given frequency response of some FIR filter, consisting of two plots Magnitude response and Phase response. - From Magnitude response plot I can easily identify number and polar angles of zeros. There are 5 zeros in my case. One of them located on the unit circle. - From Phase response plot I can see that filter is mixed phase, which means that some zeros are inside unit circle and others are outside. Here is my question: How can I find out what zeros are located inside unit circle and what are outside unit circle? Please remember that I don't have any other information about this filter besides these two plots. Thanks in advance
Unit circle and location of zeros ...
Started by ●November 4, 2004
Reply by ●November 4, 20042004-11-04
Bobby wrote:> I need to solve the following problem: > I was given frequency response of some FIR filter, consisting > of two plots Magnitude response and Phase response. > > - From Magnitude response plot I can easily identify number and > polar angles of zeros. There are 5 zeros in my case. One of them > located on the unit circle. > > - From Phase response plot I can see that filter is mixed phase, > which means that some zeros are inside unit circle and others are > outside. > > Here is my question: How can I find out what zeros are located > inside unit circle and what are outside unit circle? Please > remember that I don't have any other information about this filter > besides these two plots. > > Thanks in advanceSounds like homework -- but everything on this list sounds like homework. If I _knew_ that there were exactly 5 zeros (how?), and I _knew_ their locations, I'd look at the phase plot for the direction that the phase is breaking. Whether I did the above or not, I'd do a curve fit of the measured data. Since you know you have all zeros (how?) you can do linear regression on a_5 + a_4/e^jw + a_3/e^j2w + etc against your measured data to get the coefficients, check for goodness of fit, then take the roots of the polynomial a_5 z^5 + ... + a_0 for your zero locations. If linear regression doesn't work then you can just use zero locations directly, but finding the global minimum is up to you. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●November 5, 20042004-11-05
Yes, you are right. It is homework ...>> If I _knew_ that there were exactly 5 zeros (how?), ...<<In my case I was able easily identify that there are 5 zeros just by looking into magnitude plot. I was able clearly see at what polar angles they are. The only thing I didn't understand is whether they are inside or outside unit circle.>>and I _knew_ their locations<<If I knew their exact location I wouldn't ask this question ... Thanks Tim Wescott <tim@wescottnospamdesign.com> wrote in message news:<10okloh594mm948@corp.supernews.com>...> Bobby wrote: > > I need to solve the following problem: > > I was given frequency response of some FIR filter, consisting > > of two plots Magnitude response and Phase response. > > > > - From Magnitude response plot I can easily identify number and > > polar angles of zeros. There are 5 zeros in my case. One of them > > located on the unit circle. > > > > - From Phase response plot I can see that filter is mixed phase, > > which means that some zeros are inside unit circle and others are > > outside. > > > > Here is my question: How can I find out what zeros are located > > inside unit circle and what are outside unit circle? Please > > remember that I don't have any other information about this filter > > besides these two plots. > > > > Thanks in advance > > Sounds like homework -- but everything on this list sounds like homework. > > If I _knew_ that there were exactly 5 zeros (how?), and I _knew_ their > locations, I'd look at the phase plot for the direction that the phase > is breaking. > > Whether I did the above or not, I'd do a curve fit of the measured data. > Since you know you have all zeros (how?) you can do linear regression on > > a_5 + a_4/e^jw + a_3/e^j2w + etc > > against your measured data to get the coefficients, check for goodness > of fit, then take the roots of the polynomial a_5 z^5 + ... + a_0 for > your zero locations. If linear regression doesn't work then you can > just use zero locations directly, but finding the global minimum is up > to you.