1
$\begingroup$

I have read that "Debauchies Wavelet is simply a short FIR" so i am searching for a single FIR that i can make into a filterbank, a precise "bandpass filter" that detects narrow frequencies, i.e. 1Hz resolution and very narrow.

Having an FIR filter function allows for multithreading which is not easy with a transformation matrix, and arbitrary analysis windows.

In Reaktor DSP suite i simply used a bandpass to make an amazing filterbank program. I tried the same with a cookbook FIR filter and it was totally erratic and garbled as a filterbank analysis filter, even though it sounds good.

What FIR or wavelet can i use for a filterbank where:

 T=time, F=freq;

 for loop T++;                        // (1 second 0-44100 samples)
    for loop F++;                     // (512 bands 0-512 kHz frequencies)
       graph2D[T,F] = FIRfilter(T , F );

EDIT- on request, schematic of a bandpass that can be used for audio analysis: enter image description here

$\endgroup$
7
  • $\begingroup$ Alas, narrow band pass filtering (relative to the sample rate) require long FIR filters, not short ones, especially if you want high stop band attenuation (for which you did not specify a value in your question). $\endgroup$
    – hotpaw2
    Commented Mar 28, 2016 at 12:21
  • $\begingroup$ Thankyou Hotpaw, The following image is what i have previously programmed with a bandpass in Reaktor, i simply want to find a similar filter in code, the resonance was about 0.97, which meant that the filter amplified massively any amplitudes of a particular signal that was detected at precisely the designated frequency: en.wikipedia.org/wiki/Wigner_distribution_function#/media/… $\endgroup$ Commented Mar 28, 2016 at 18:14
  • $\begingroup$ Have you measured the bandwidth and stopband attenuation of that filter? $\endgroup$
    – hotpaw2
    Commented Mar 28, 2016 at 19:25
  • $\begingroup$ Hi, i was suprised to find that i had used a Reaktor5 integral "2 pole bandpass", so i have added a shematic of it to the original post, it's made of a "BiQuad I" and a "SetBQPB" element. I had measured the response, it's similar to a perfect guassian bell curve where the apex narrowness is adjustable by change the resonance, used at .95, which creates a trade off between time ant frequency resolution. It's just a perfect bandpass, it's difficult to measure precisely. $\endgroup$ Commented Mar 29, 2016 at 0:32
  • $\begingroup$ A biquad is an IIR filter. So you want a FIR filter with a response that approximates a given IIR filter? There are already other Q&A's here on that. $\endgroup$
    – hotpaw2
    Commented Mar 29, 2016 at 1:24

0

Your Answer

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