Alternative to unwrap()

[Børge Strand-Bergesen]

I'm working with Fourier Transforms. In my application,
discontinuities add a lot of problems. (I'm writing a program that
compares the gain and delay responses of a computer generated filter
to those preferred by a user.) I wouldn't know which journals to start
reading to improve on unwrap().

One idea I have been toying with is to add n*pi to the angle of each
polar coordinate instead of n*2pi (for integer n). Whenever n is odd,
I compensate by negating the absolute value. This "bipolar absolute"
actually works very well, and I'm able to extract continuous phase and
delay plots. Unfortunately, my method is a bit cumbersome and a far
cry away from a one-size-fits-all. I need to be confident that the
method works for a wide range of filters. Eventually it will have to
be optimized for high-speed paralell execution.

So any pointers to research on clever unwrap functions would be
greatly appreciated.


Thanks,
Borge