I have some intuition about bandwidths from these videos (1 and 2) explaining how the frequency response corresponds to the Z-surface.

The only thing I could find on the internet about calculating the bandwidths is this matlab tutorial:

The bandwidths of the formants are represented by the distance of the prediction polynomial zeros from the unit circle.

[frqs,indices] = sort(angz.*(Fs/(2*pi)));
bw = -1/2*(Fs/(2*pi))*log(abs(rts(indices)));

but bw is not the distance to the unit circle but the log of the norm.

This is assuming the roots/poles are inside the unit circle or bw will be negative. But I'm unclear why the roots/poles shouldn't be outside the unit circle?

I can see that the relationship of the bandwidth is not linear with the root/pole position (relative to the unit circle), but why is it logarithmic?

Perhaps, if the root/pole is outside of the unit circle we should reflect log(x) around 1 i.e. log(2-x):

bw = -1/2*(Fs/(2*pi))*log(2-abs(rts(indices)));


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.