| previous by date | index | next by date |
| topic list | next in topic |
Some time ago, I sketched out a plan for a series of such sub-modules that could be patched together to create vocoders, Shepard function generators, and other such "spectral" effects. Fully modular (like the "SuperMoe for Vocoders). I don't have my notes with me, but I think it was not many modules... filterbank, envelope follower bank, VCA bank, 8-phase LFO... I can't remember now.
Sounds like heavy hardware, but my thinking was that the VCAs, etc, need not be super-high quality. The multiphase LFO I was thinking of doing cheaply ala Simonton's design for the PAIA Shepard function generator. The EF bank I was going to use companding chips, the way that Anderton did, to keep parts counts down. I wanted the money to go into more bands, so I was going to compromise on the little details, since these were very specific modules and not meant to replace 8 MOTM VCAs or anything like that for critical aspects of a voice (although they would be available individually).
Graphic equalizers tend to have octave spacing, but I think for this kind of work, it's not always best. For instance, Serge's FFB uses "7th" spacing so that, over the audible spectrum, certain tones and their harmonics aren't over-emphasized.
I was actually playing around with the idea of selecting the center frequencies of the bands based on key human vocal formants. You know, target ∗those∗ to improve vocoder response, rather than arbitrarily spacing them according to the octave. Of course, this favors human speech, but somewhere I had found a compromise that I thought would work well generally. The bands WERE equally spaced, yet fell quite close to male & female key formants. It was like... 1/5.5 octave spacing or something odd. Looked good on paper...don't know if it's a fool's errand or not.
Wish I had my notes. It was a huge, deeply considered plan. Just on paper, of course.
Mr. T
-----Original Message-----
From: paulhaneberg [mailto:phaneber@...]
Here's my ideas for a FFB. 18 or 20 bands, 1 octave apart for the
highest and lowest bands, 1/3 octave or less apart in the middle.
Individual outputs for each band and an envelope follower for each
band as well with the follower response tuned to that particular
frequency range. A follower output for each band as well.
This thing could be used as a front end for a vocoder. A second one
with the addition of a bunch of VCAs could be the back end of a
vocoder. You could split a sound into a bunch of bands and
manipulate each band differently.
I'd love to hear others ideas on this thing as well.