fffb~ Q-factor confusion - building a 1/3 octave bank

    EMV
    Feb 21 2010 | 5:16 pm
    Hi all,
    I have a question about fffb~, and the Q-factor in particular. I'm building a third-octave filterbank and a whole-octave bank using fffb~.
    I based my patch on the "fffb~ third-octave EQ" from the examples.
    There they use a Q-factor of 0.793701, but if my math is correct (Q = f(0) / delta f), that would mean a bandwidth of 1259 Hz for the filter at 1000 Hz, which is equivalent to 1.7149 octaves.
    That's a bit much for a third-octave filter isn't it?
    According to http://www.sengpielaudio.com/calculator-bandwidth.htm, for a 1/3 octave filter with crossover points of -3 dB, the Q-factor would have to be 4.3185
    So I'm confused! Why did the max/msp example take such a low Q (high bandwidth)? Can someone confirm that the Q-factor of 4.3185 is the one I need?
    (added for convenience: the patch from the examples section)
    • EMV
      May 07 2010 | 12:03 pm
      Nobody?
    • j.y.bernier
      Nov 19 2011 | 5:55 am
      In case you are still interested in the answer:
      0.7937 is quite ridiculous for a third-octave EQ. With fffb~, you should use around 9 to get -3dB at cutoff. The drawback is that you get some ripple. Peter Elsea uses 10:
      ftp://arts.ucsc.edu/pub/ems/maxtutors/MSP.tutorials/ThirdOct.pdf
      Also, you may use negative gains fot a boost/cut EQ, although cut and boost profiles won't be symmetrical. Overall, a decent +/- 24dB EQ can be implemented with fffb~. A gain maping function is needed.