fzero~frequency limit @ 2500 Hz

    Jun 23 2018 | 7:21 am
    For the first time I use fzero ~ (since the excellent fidle ~ and sigmund ~ no longer work), and I find that it does not allow analysis above 2500 Hz. And more: testing with a ~ cycle, at 2000 Hz tell me that we are around 1000.
    Does anyone know anything about it?

    • Jun 23 2018 | 8:49 am
      Hi Stefano, I don't know about your problem, but here you can find 64-bit versions of fiddle~ and sigmund~: http://vboehm.net/downloads/
    • Jun 25 2018 | 6:54 am
      Thank you very mutch, it's a great news.
    • Jun 25 2018 | 10:59 am
      2000Hz seems to be working fine for me here:
      Max Patcher
      In Max, select New From Clipboard.
      Were you testing differently?
      2500Hz is a real limit, though. The fzero algorithm trades good low frequency performance for poor high frequency performance.
    • Jun 27 2018 | 2:38 pm
      You might be able to get around that by upsampling it inside a poly~. Away from my computer at the moment so can’t test this, but if 2500 is represented within the algorithm as a fraction of the samplerate, increasing the samplerate should raise the cutoff point. You can test this by simply increasing your samplerate. If that works, wrap fzero~ in a poly~ Yourpatchname up 2.
    • Jul 01 2018 | 7:56 am
      Yes, I agree. In any case the new fiddle~ and sigmund~ are better. Thank
    • Nov 11 2018 | 4:53 pm
      FWIW, the upper limit for fzero~ in Max 8 has been upped to about 8800 Hz.
    • Nov 11 2018 | 7:02 pm
      Well, better, but not very good.
    • Nov 11 2018 | 7:16 pm
      More specifically, the max frequency is set to the sample rate * 0.2. So, higher frequencies are possible at higher sample rates. Realistically, the algorithm probably needs at least 10 samples in a period to work properly. 8820 Hz at 44.1 kHz sample rate is 5 samples, which is really pushing it.
      I think that's nearly the top of useable fundamentals for audible music. I can imagine wanting to identify other fundamental frequencies, but I think that these are a bit of an edge case? At that point, the hottest fft bin is always the correct answer because there are no salient harmonics.