On Thursday, December 5, 2013 at 9:21 AM, vallste@libero.it wrote:
Dear List,
I'd like to remove some frequencies zones (a band reject filter) from a signal with very precise frequency bands (high rolloff?). My Idea is somehow the following (Butterworth filters just for reference):
ainputSignal;Incoming signal
ainputSignalB1 butlp ainputSignal, kfreq1 ;Band1
ainputSignalB2 butbp ainputSignal, kfreq2, kband2 ;Band2
ainputSignalB3 butbp ainputSignal, kfreq3, kband3 ;Band3
ainputSignalB4 buthp ainputSignal, kfreq4 ;Band4
then the single bands are processed (or not) indipendently; If I modify only one band I can remove (from the input signal) only the band I want modify and sum the band once processed while keeping the rest of the signal intact. For example if I process only Band2:
aoutputSignal = ainputSignal - ainputSignalB2 + ainputSignalB2Pocessed
Does this make sense or there are some phase modifications (in the filtered signals) to take in account?
Then, which filters (or combination of, or pvs tricks, or...) to use (instead of the Butterworth written above) to have a very steep rolloff but being able to run the thing in realtime (on an average machine, not a supercomputer nor a pentim 2)?
Ciao
Stefano