Why I don’t particularly like spectral (FFT) processing

1) Smearing of transients

2) Disproportionate favoring of the highest frequencies (assuming a sample rate of 44100hz, half of the FFT bins will be dedicated to the highest audible octave)

Just those two things alone are usually enough to make FFT-based processings have a sonic character that will be all over the processed sounds. Which could be just fine if one happens to want that kind of sound, but then, there’s no obvious way to get rid of that character…

I dunno, would looking into multiresolution-FFT be some solution to this? Or haven’t I just bumped into really good apps or plugins that do FFT-based processings?

This was just a random rant.

Advertisements
This entry was posted in Uncategorized. Bookmark the permalink.

One Response to Why I don’t particularly like spectral (FFT) processing

  1. the finger, and the mouth says:

    multiresolution fft seems like an advance to me (e.g. izotope rx). i perceive the general type of
    artifact that results on transients as “blurring” or “diffusion”, as compared to the “smearing” and
    “ringing” effects on transients that i associate with iir and fir filtering. purists may claim that any
    pre-response is unnatural. pre-response only occurs in nature when some of the sound is
    transmitted through a solid medium, such as a wood floor, and this would be mostly low
    frequencies… so maybe that is the crux of your problem with fft. i myself find the artifacts
    of fir (ringing around the transients) to be as unpleasant as the “softening” effect of fft
    can be annoying. in mastering i preferred to use all three types of filters together and
    tune them meticulously by ear, so reducing and attempting to cover all the artifacts of
    all types of filtering, unless they help with the illusion, iir smearing is a natural effect, so
    controlling this seems importatnt too, which leads us to the linear phase types and their
    artifacts… there is no fourth option for filter types afaik, or will you make a
    “granular filter” that is linear and can be calibrated?

    note that since the bulk of a snare drum hit is pretty much pure white noise, any fft process
    performed on such material may produce slightly non-linear results, and it would be questionable
    whether that non-linearity can be audible, but perhaps the artifacts of an fft process could
    indeed help to color such seemingly boring (to me) “uniform” sounds…so when the options to smear
    or add ringing are not desirable, doesn’t fft seem interesting? as you mention, it
    cannot do low frequencies… that is interesting to me since it opens the door for other processes
    which have an exaggerated effect on bass… ultimately it can lead to more control…

    however, the idea to combine different processes and adjust them completely by ear… as if everything were
    slightly non-linear and results are highly program dependent seems to throw a wrench into the idea that we
    should have to work less, not more to get our music made. and “non-linear
    seems to be itching powder for science, simply because science is based upon being
    able to predict responses… well too bad, this is art… someday, i think it will come to that.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s