spectral reverser

    Oct 03 2010 | 11:18 am
    Hi guys out there, trying to build a spectral reverser, basically it take a spectrum and flips it on the frequency basis, highs become lows and the other way around. Was thinking of usinf pfft~, any simple idea how to do this?

    • Oct 03 2010 | 3:50 pm
      i did that for an installation in amsterdam at sonic acts this year http://www.steim.org/STEIMBLOG/?p=1481 (scroll down to "two rooms, flipped) the route i took was millers sigmund~ object and then resynthesize with the frequencies flipped. caveat: due to the logarithmic nature of the frequency perception, you have to tweak the output mathematically, otherwise you end up with just a high pitched hiss. its a fun endeavour anyway.
      hth hans
    • Oct 04 2010 | 9:50 am
      In my opinion you have to la' a spandi ledam. Check if it is working...Good luck.
    • Oct 04 2010 | 5:56 pm
      now thats advanced technology :-) elisabatti could you explain a bit more, please?
      thanks hans
    • Oct 06 2010 | 1:58 am
      if i understand well bin 0 transformed to bin 511, bin 1 to 510 etc... i have something like this...
      the main idea is to store the input bins and then index them with the way you want (backwards)..
      but don t expect something awesome..
    • Oct 08 2010 | 4:11 pm
      Well Tada, i get only noise form your patch, should look at it now.. on the other hand Hans' idea works fine and it's rather easy.. one thing Hans, did you remap logarithmically the frequencies? which kind of tweaking did you find best? I guess it depends also on the material you use...
    • Oct 11 2010 | 3:36 pm
      hi jester,
      yes, i used a logarithmic mapping based on lp.scampf (litter starter pack) and manually tweaked it until it sounded good for my purpose. (as a reference soundfile i used a snippet of the poplvetsian dances by borodin. easy to hear leaps in the melody and then have them jump in the opposite direction while retaining some of the soundqualities).
      i also wasn´t able to get much out of tadas example (which otherwise seems like a very elegant approach). maybe have to study it more carefully.
    • Oct 11 2010 | 11:58 pm
      hey... ive been reading you are working with the spectrums... do you know any patch that can "read" images and reproduce them as sound? thanks a lot. martín
    • Oct 12 2010 | 7:36 am
      hi martin,
      actually i am not sure i understand what you want. image to sound: either openraw in soundfile or search in the jitter forum, loads of posts about that.
      or did you have somethng specifc in mind using spectrums?
    • Oct 17 2010 | 10:58 pm
      Hi hans, sorry for taking so much time to answer but i had to work far away from the computer jej. Thanks very much for your answer.
      I´ve been looking for that in jitters´forum but i couldnt find any solution.
      In order to leave things clear, i want "something" to read this scaned image of a "thread" (a thread used for packaging the pizza jej) like if it was a sound wave (check the attachment).
      Any ideas? I have found one that can record sound, show the spectrum and reproduce that spectrum in loop. but it doesnt work for "long-lasting" images nor for other non-spectrum iamges.
      Nice week! Martin
    • Oct 19 2010 | 7:47 am
      Your question is confusing. You want to translate this image into a waveform, right? Check [jit.buffer~] (of course, you'll need to know about Jitter, matrices, and stuff like that). But you talk about spectrum: this is a different subject. As far as I understand, you don't want to work in spectral domain, rather play back a waveform, right?
    • Oct 19 2010 | 11:53 am
      What you could try is to build a vocoder and reverse te connection between the analysing and modulating part. The synton Vocoder could do this with a matrix. I tried this and it works fine. Also, you have more control since you can manipulate the matrix to alter the sound to your liking.
      Not FFT but a workable solution.
    • Oct 19 2010 | 10:46 pm
      i think elmarta is looking for an image to spectrogram patch, where the frequency bins are mapped on the y-axis, the color determines the strength of each frequency bin (the z axis) and time is the x-axis. I saw MATLAB code that did this once and it was awesome, but don't have the code myself. Would love to see a max version. Could try to code it myself when I have a bit more free time.
    • Oct 25 2010 | 1:25 am
      Hey.. thanks for the answers :) As you can see Im new at this.
      Its good for me that my question has been interpretated in different ways so that it gives me some chance to imagine how to solve it.
      I want to playback a waveform, as Jean-Francois Charles said (thanks for your answer, i ve read your article and downloaded the patches). The x axis=time, y=frequence, as ComfortableInClouds said but Z=¿? (there is no color variable though its a scan of a white thread commonly used to wind up pizza box).
      Danny im sorry but I dont know how to use matrices. I was thinking on this process.. import an image, read the contour )using black and white contrast) and reproduce it as if it was a waveform. Some variables i want to manage are: length and height of the imported image so to manage duration and frequency in the resulting sound. Here im trying to assume some parallelism between distance and time (1cm to 1sec); and height and frequency.
      Any tutorial from where i can learn how to start doing this? Thanks a lot for the time you take to understand what im trying to say and for the help :)
      Again, Nice week jeje Martin
    • Oct 03 2013 | 12:11 am
      has anyone implemented a spectral reverse with the logarithmic curve, not using third-party externals? i'm not entirely sure how to do this..
    • Oct 24 2013 | 9:01 am
      I caved in and started using tap.spectra~ - but I'd really like to implement a log. response in it- any ideas?