[sdiy] For you FM synthesis fans

Richie Burnett rburnett at richieburnett.co.uk
Thu Nov 20 19:25:40 CET 2014

You can do reasonably good formant synthesis using FM modulator/carrier pairs. So if you can extract the pitch, formant centres and formant widths from the spectrum of a sound, it's not that difficult to achieve the target spectrum given enough operators.


Sent from my Xperia SP on O2

---- Scott Nordlund wrote ----

>> Thanks for the feedback, Scott!
>> As for the rumors, the story I heard some decades ago (I think I even
>> read it in the German issue of Keyboards magazine, but I would have do
>> dig deeply to verify that) was that at least one particular sound
>> resembling a plucked wire of the Golden Gate bridge was first modeled
>> using Karplus-Strong, then analyzed and transformed into according DX7
>> parameters.
>> Rainer
>My guess is that at most they could attempt to extract an amplitude envelope, a rough estimate of spectral bandwidth (modulator envelope), and maybe some basic information about the harmonics present, which could be used to select carrier and modulator frequencies. But that's only a very crude and incomplete approximation and really just a starting point for programming a sound. I don't think it counts to call that resynthesis unless you've only got two operators.
>Additive resynthesis was being explored already at Bell Labs in the mid 60s, and reached a pretty good level of sophistication in the 80s (PARSHL at Stanford). Harmonic additive resynthesis is pretty simple, and could be done by the Synclavier in the mid 80s, so it's not that big a deal to make something that automatically programs a Kawai K5000. But FM isn't nearly as straightforward. 
>Synth-diy mailing list
>Synth-diy at synth-diy.org

More information about the Synth-diy mailing list