No announcement yet.

Cleaning up audio streams

  • Filter
  • Time
  • Show
Clear All
new posts

  • Cleaning up audio streams

    So I've finally figured out how to capture iPlayer (Flash) streams, just in time to catch some Richard Strauss pieces I've had on my list. Also just in time to buy 8, but I think the stuff here is more or less version-independent. Wonder if somebody has other thoughts on this.
    • The low end (very low - below 15Hz) usually seems to have a sharp uptick viewed in the Spectrum Analyzer (DC7), sometimes to high enough levels (-20 or even above) that it interferes with other processing. Requires multiple applications of a high-pass filter at 20 Hz and 24 db/octave. For music, I prefer the Butterworth characteristic. I've tried the brick-wall filter but it doesn't seem to work well at low frequencies - maybe I need to adjust something? I don't want to use something that affects content much above 20 Hz.
    • BBC's stream is good up to a very sharp 15 kHz cutoff - sort of. There are a couple of little peaks above that, like a very attenuated FM stereo pilot, but usually in the 17-17.5 kHz neighborhood. They're generally below a CD's noise floor, but can get into the -75-80 range when there's major low-frequency content. Are these Flash artifacts and is it worth the effort to whack them (the notch filter helps).
    • Sometimes they have a low-level but noticeable buzz. 50Hz notch didn't help much, but the Harmonic Filter for 50Hz buzz (just the basic), even harmonics, and only 20-25 dB attenuation seemed to do a good job of making it inaudible without hurting the rest of the sound much. Interesting that odd harmonics didn't seem to be present. Source fluorescent lights in the booth or ??
    So given all that, my workflow ended up simple: record; trim ends & fade in/out; 24 db/oct bandpass at 20 & 18; notch out the remaining high peaks if they get above -90 or so; listen for buzz & fix if present; tip the high end (above 10 kHz) up just a tiny bit with the 20-band; another bandwidth this time at 15 and 18; and save. Close, reopen, and break into cuts and/or make MP3. BTW, I always swore by the classic mode but have now come to love fast-edit...

    One other question: if I use these in the car there's a problem -- BBC orchestral (BBC 3) streams have about a 50 dB dynamic range in the music - appears that they don't use any compressors. So I need to compress car versions. What's the best setting for that? What might approximate, for instance, a FM station that's not going for the "loudest" crown but still wants to be heard?

    -Mike B
    Last edited by Craig Maier; 05-27-2019, 04:31 PM.

  • #2

    I think that the low frequency uptick seen on any fft based spectrum analyzer is the result of the fact that the bottom band can not goto zero hertz. In other words, there is no such thing as direct current (DC) unless integrated over a time interval from minus infinity to plus infinity which can not happen in the real world. The steadiest signal had a start from some value and ends at a different value, ultimately. So, fft analyzers will always show the bottom band (the resolution of the analyzer) as an uptick. To see what I am saying, raise the fft size of the analyzer and you will see a down-tick in the frequency of the lowest band up-tick.

    I do not know about the BBC's response above 15 kHz; I have not looked at that.

    Could the buzz be comming from your computers mother board getting into the soundcard? That would create a variable frequency situation more amenable to CNF type noise reduction. But, not really sure.

    Last edited by Craig Maier; 02-11-2014, 10:11 PM.
    "Who put orange juice in my orange juice?" - - - William Claude Dukenfield


    • #3
      Thanks. I hadn't thought of trying different FFT sizes - though if I go much above the standard (4K 1 think) response drastically slows down on my computer. Will check. But it sounds like a measurement artifact that's not actually in the file - good.

      Did a couple more tonight - the biggest high-end spike actually is just about at the frequency where the cutoff begins - so I suspect artifact of some filter. The notch filter brings it under control nicely. I've also stopped worrying about whatever I see above that since only rarely does it spike over -100, so it wouldn't even reach the noise floor of a CD.

      The buzz is file-specific. Most don't have it; a couple did. The 50Hz harmonic filter with even harmonics did an excellent job on it.


      • #4
        Ok - especially on your last point re 50 Hz buzz. Be sure to use the smallest number of harmonics to do the trick so as to minimize damage to the audio itself. Obviously, since it is file related, it has noting to do with your system.

        "Who put orange juice in my orange juice?" - - - William Claude Dukenfield