interpolating filtergraph's filtertypes

    johannotto
    Sep 13 2008 | 5:33 pm
    hi there, I am trying to interpolate between different filtergraphs-filtertypes with a textobjects feeding another filtergraph with coefficients. it works - but not very well, because during interpolation, the first filtergraphs-params (freq, type - pattr aware) are banging the filtergraph-object to output its "non-interpolatet type" again... and interpolation is gone... perhaps somebody has a more smarter/simpler solution? here's the test patch:
    • oliviertouche
      Sep 14 2008 | 10:47 am
      Hi, if I understand well, you should try and have a look a the subpatcher called [p cascade] in the filtergraph's help patch.
    • johannotto
      Sep 14 2008 | 11:52 am
    • Gregory Taylor
      Sep 14 2008 | 12:06 pm
      Since the svf~ object outputs the results of lowpass/highpass/bandpass/notch filtering simultaneously, why not use *it* and then pattrize the *~ objects that handle the output amplitudes for each filter output type? The only thing there that might require rubbing a few neurons together would be setting up the three/fout-way crossfade [and no, there isn't a single Max object that does N-way fades]. It's quick and dirty and sounds just fine.
    • johannotto
      Sep 14 2008 | 12:25 pm
      hi, thx gregory. I also thought on this possibility (and I've also build up a 5way crossfader for the wavesources)... Using a svf~ with just graphical-feedback of filtergraph should be ok, does it?
      By the way 2up-svf~ isn't able to output it's filterouts in parallel, or am I wrong?
      johann
    • Gregory Taylor
      Sep 14 2008 | 7:10 pm
      I don't use the filtergraph~ stuff unless I've got no other choice, preferring to *listen* to the filters and to map the three or four faders' worth of stuff I'd use so it's under my fingers instead of eye candy on the screen. Your mileage may vary.
      The plain old MSP svf~ object kicks out parallel outputs from a single audio input - see the help file - so it's perfect for morphery. Of course, if you cannot live without eye candy, you can use the more um... exotic outputs from the filtergraph~ object to set things, although they won't be wildly useful when morphing filtertypes.
    • johannotto
      Sep 14 2008 | 7:23 pm
      yea, you're right. I also once compared between filtergraph~/biquad and svf~ - liked svf~ much more... just get to much used to the graph-reference... it's time to stop that now ;-) thx!
      by the way: the 2up_svf~ Version doens't support parallel filtertype-output
    • Stefan Tiedje
      Sep 23 2008 | 7:27 am
      johannotto schrieb: > perhaps somebody has a more smarter/simpler solution? here's the test > patch:
      Answering late, but the original question had not been answered:
      Instead of the filter graph, store the coeffients in a pattr. Problem, if you feed that back to the filtergraph for display, it will switch to display mode and you can't play with it easily anymore.
      I guess Jeremy choose to store the type/freq/gain/Q values of a filtergraph~ instead of the coefficients for this reason. In most cases you would want to stay in edit mode if working with filtergraph.
      My solution would use a second filtergraph~ only for displaying purposes, but you can use transparency and "ignore clicks" to make it into a nice userinterface. I switched on also "imitate analog flavour" to make it sound better... I'd wonder how it compares sound wise with svf~, for sure its way more flexible...
      You could by the way also include the type/freq/gain/Q values into pattr, but exclude them from interpolation, that way they would jump in at the fixed preset points...
      Stefan
      -- Stefan Tiedje------------x------- --_____-----------|-------------- --(_|_ ----|-----|-----()------- -- _|_)----|-----()-------------- ----------()--------www.ccmix.com