Feedback Problem in Bandpass-Setting


    Feb 19 2006 | 8:58 pm
    Hi list,
    i have quite a problem:
    Im working on a piece, where i use a bank of bandpass filters (four
    parallel cascades of two biquads at gain 1 and Q bout 15 on every one
    of the four input channels) to loosely imitate the resonances of an
    instrument. The output of the filter should then be reverbed a lot to
    generate a cloud of airy sound. Here seems to be the problem: Reverbing
    the filteroutput more or less amplifies it, as it is present over
    longer a time and gets accumulated. Thus the whole thing is extremely
    prone to feeding back into the filter...
    I've tried building feedbackdestroyers with fft, but when those shut
    enough to kill the feedback, i cant get anything else past them into
    the filter. i've tried working on the reverb settings (used newverb) as
    well, but when i got the feedback under control, the reverb level was
    almost impercrptible over the direct sound from the instruments -
    timpani i forgot to mention...
    Could it just be a problem of the very unfavorable accoustics of the
    room in which i' m working, - its small and the instruments are in a
    corner, so i get a lot of reflection from the walls and the ceiling,
    and the huge membranes reflect everythig into the mics. Its not just a
    problem of the newverb, if tried other devices as well...
    Any ideas how to get a sort of tuneable resonator/filter bank without
    the self exitation problem?
    i'll try to get an isolated patch without any externals missing for
    posting...
    Titus Bellwald

    • Feb 19 2006 | 9:59 pm
      Something you might try is doubling the number of filters, and reducing
      the Q. This will make them be less prone to ringing. Also, since
      you're using the reverb to amplify the filters, you might also try
      reducing the gain.
      i.e.
      input
      |
      reson~ 1. 400. 7.5
      |
      reson~ 1. 400. 7.5
      |
      output
      Also, have you considering routing the audio in the opposite manner
      (reverb, then EQ)? Depends on how you're using it, so this might not
      work stylistically...
      Additionally, I'd consider a compressor/limiter on the input and output
      stages to manage volume levels.
      Peter McCulloch
    • Feb 20 2006 | 7:12 am
      I'm not fully sure what you are doing, but I get the impression that
      you are micing the timpani, processing and then play back using
      loudspeakers. If so, the first thing I would consider is the distance
      and position of mic and loudspeakers, and also what kind of mics you
      are using. Sorry if these suggestions are to basic.
      Another thing to try is to let the center frequency of the filters
      drift a little bit. This might prevent them from being triggered at
      the exact same frequency.
      One more thing you could try is introducing freqshift~ to shift
      frequency downward by a few Hz. This is a common way of avoiding
      feedback, as any frequency that starts building up through feedback
      will be pushed downward frequency-wise and hence gradually pushed
      away from of the resonant frequency. It might introduce unwanted
      detuning and artifacts though.
      It's well worth checking out if you might benefit from limiters or
      compressors as well.
      Best,
      Trond
    • Feb 20 2006 | 9:12 pm
      Titus Bellwald wrote:
      > Hi list,
      > i have quite a problem:
      > Im working on a piece, where i use a bank of bandpass filters (four
      > parallel cascades of two biquads at gain 1 and Q bout 15 on every one of
      > the four input channels) to loosely imitate the resonances of an
      > instrument. The output of the filter should then be reverbed a lot to
      > generate a cloud of airy sound.
      You might consider to combine the two things. As the resonance of an
      instrument is in general a simplified reverb, you could build the reverb
      from scratch and ignore the main problem of reverbs, the avoidance of
      resonances. Thus its easier to build (less delay lines) and you get your
      resonances as well. It requires some math to find the reflection length
      aka delay times to get your resonances. You could still put some filters
      into the feedback, but these could be simple onepole~s.
      Just a suggestion...
      Stefan
      --
      [][] [][][] [][] [][][]
      [][][][][][][][][][][][][][][]
      Stefan Tiedje
      Klanggestalter
      Electronic Composition
      &
      Improvisation
      /~~~~~
      \ /|() ()|
      ))))) )| | |( \
      /// _/)/ )))))
      ___/ ///
      -------------------------x----
      --_____-----------|-----------
      --(_|_ ----|-----|-----()----
      -- _|_)----|-----()-----------
      ----------()------------x-----
      14, Av. Pr. Franklin Roosevelt,
      94320 Thiais, France
      Phone at CCMIX +33-1-49 77 51 72