synthesizing/modeling reverb?
i'm wondering about techniques for synthesizing reverb as part of the sound synthesis itself. like as a "physical modeling" type of thing, not something you add after as an effect. kinda like sympathetic resonance modeling on a larger scale. :)
i haven't found much info searching, "reverb synthesis" is pretty useless. :) what i've read about sympathetic resonance has been pretty helpful, but i haven't found much info either. i think i need some better terms to search for that i don't know. i'd appreciate any pointers. :)
it's funny, i find that if you use the "olden days" modular/multitrack techniques of synthesizing reverbs, like reusing your envelopes from the sound parameters, feeding them back, parallel processing, layering you can get some pretty wicked "fake" reverb sounds that i haven't heard in after the fact reverb effects. i'd guess because those reverbs work linearly and can't effect the original sound that goes into the reverb?
hmm, maybe some kinda dynamic convolution of synthesized resonances....or i'm going to have to hear what happens if you continuously reconvolve the resulting sound with the previous version.. a sine wave? ...kinda funny to think of a physical space going back into/outside itself.
anyway, when you think about it, with physical instruments, because reality and laws of physics :) you have a pretty good idea of what effect the space will have on a sound before it happens it and that will influence the sound you choose to use and how you articulate your playing of the instrument intuitively. or if there is a certain musical effect people have wanted to amplify they build the space around the usage of certain instruments. closer feedback loops....
i believe that a room reverb can never be "part" of the synthesis process itself, it is by definition something which is applied as last step, i.e. as an individual process.
but what can be very useful in the mid of synthesis processes are basic resonance bodies (think guitar), for example built with comb filters.
sawtooth oscillator or karplus/strong model --> fffb~ or resonators~ --> FIR filter, harmonic exciter or even FM.
-110
>i believe that a room reverb can never be "part" of the synthesis process itself
I believe you speak too soon, old CX.
Using a full-on "normal reverb" as part of the loop in karplus-strong sounds really amazing.
thats true, a reverb is one of the few things which almost always works in a feedback loop (given that there is some kind of phase diffusion in the algo)
hard to say where the border between post-effect and part-of-the-synthesis is here.
just off the top of my head, everything you’ve mentioned that you’re interested in, reminds me of ‘waveguide synthesis’(on the resonance end of things), and ‘feedback delay networks’(on the reverb end of things).
you could do searches on those two terms and maybe combine them somehow…
yeah, i think you're right. and yeah they're both ideas i like to mess with as well... i guess i was being kinda lazy. :) just need to work to integrate them more. it's funny to me because of all the "waveguide/physical model" stuff i've used/made, none that i can recall sounded like the instrument was in a completely "dead" acoustic space, so to me they feel naturally linked.
Using a full-on "normal reverb" as part of the loop in karplus-strong sounds really amazing.
haha. yes, it's fun. :) i'd really like to try a quantec yardstick in a feedback loop to hear/know more about what it's doing, at least, but it'll be a while before i can afford one! it has spaces like bread box and circus tent! :O their faq has some interesting information. http://www.quantec.com/index.php?id=faq041 this one actually gives me a pretty good guess as to at least some of what's going on.
anyway, i accidentally :) found this paper http://msp.ucsd.edu/Publications/icmc11.pdf looking up non-linear reverbs. striking a physical object attached to a contact mic and then using linear predictive analysis to remove the natural resonances and process the sound in a feedback network...i've done similar to this, but i didn't know about about "linear predictive analysis"! :) it's funny, it actually is pretty cool sounding anyway without removing the resonances and also to take your processed sound and send that back out into a contact mic to re-vibrate your physical object...
but that gives me a few ideas to work backwards from. :) gonna start messing with lpcToolkit and rtcmix, i guess.