poly~ upsampling methods


    Apr 08 2007 | 7:25 pm
    I just noticed that if I have a poly patch within a parent patch that downsamples by two, it's output is simply the downsampled values repeated.
    Ie the sequence 1 2 3 4 5 becomes 1 1 3 3 5 5.
    Is there a way to specify how the output of poly is upsampled to produce its output in the parent patch say to get something like 1 0 3 0 5 0 (ie zero padding). Or possible interpolating...
    I seem to remember talking with somebody about this a while back but I can't seem to find any info in the forum.

    • Apr 08 2007 | 8:14 pm
      weird, i think poly~ has its own signal vector? check the documentation.
    • Apr 09 2007 | 3:31 am
      Ok I see what your saying. When you downsample by 2 within poly the signal vector size becomes half of that of the parent patch. I made a workaround by using a count~ to multiply the output of the poly~ so that every second sample is set to zero.
    • Apr 09 2007 | 4:36 pm
      Quote: Tyler Nitsch wrote on Sun, 08 April 2007 21:31
      ----------------------------------------------------
      > Ok I see what your saying. When you downsample by 2 within poly the signal vector size becomes half of that of the parent patch. I made a workaround by using a count~ to multiply the output of the poly~ so that every second sample is set to zero.
      ----------------------------------------------------
      you would have to interpolate it yourself - outside the poly
      patcher.
      it was about 2 years ago here in the list where someone
      gave an example, but i cant find it anymore. it was something
      using count~ so that you know where you are ...
    • Apr 13 2007 | 4:33 pm
      Tyler Nitsch schrieb:
      > Ie the sequence 1 2 3 4 5 becomes 1 1 3 3 5 5.
      which is what you would have to expect, you throw away information...
      > Is there a way to specify how the output of poly is upsampled to
      > porduce its output in the parent patch say to get something like 1 0
      > 3 0 5 0 (ie zero padding). Or possible interpolating...
      zero padding would screw your signal completely, why would you want that?...
      Interpolating is just a low pass filter of half the samplerate (outside
      of the poly~)...
      Stefan
      --
      Stefan Tiedje------------x-------
      --_____-----------|--------------
      --(_|_ ----|-----|-----()-------
      -- _|_)----|-----()--------------
      ----------()--------www.ccmix.com
    • Apr 13 2007 | 7:02 pm
      Hey Stefan,
      In a lot of DSP literature they all deal with upsampling thru zero insertion. I was trying to develope a multirate system that was based off of the perfect reconstruction filters. I used haar filters. The idea is to split a signals spectrum in half with the analysis filters... downsample each half and then upsample and filter using the resynthesis filters. Basically a one level realtime wavelet system. When you repeat sample values in the system I created it increases the overall power of the resultant signal which rendered the resythesis part of the patch inaccurate... after doing the zero insertion it worked. Unfortuneatly I cant' get more than one stage to work properly....