[sdiy] Resonator type filters

[Tom Wiltshire]

At Ian Fritz's suggestion, I've been reading the following paper:

M. V. Hathews and J. Kohut, "Electronic Simulation of Violin Resonances", jj. Acoust-Soc. Am. til, IbHJ (1973).

It's not available anywhere for free, as far as I can tell, so you'll need a friend at a university. But I'm going to tell you the gist of it.

When I hear "Resonator type filters", I'm thinking of "Body modelling filters" - e.g. filters that allow you to model the body (and other) resonances in acoustic instruments to either synthesise them more accurately, or to add another dimension to other synth sounds that may not be a copy of a particular acoustic instrument, but which might have a particular character ("woodwind" say).

What is required for this varies a fair bit. What I knew already was about brass, woodwind, and vocal resonances, which are fairly few in number (5 would cover it - 4 is ok, even 3 does a reasonable job, as Elhardt's PolyMoog video demonstrates). You need less for brass and woodwind and more for vocals.

What I've learned from the paper posted above is that for strings this is not sufficient, although there might be a couple of major resonances lower down that you could copy fairly easily. But the "lush", "rich" quality which we're after seems to come from the amplitude modulation of the harmonics caused by many narrow bands coupled with a limited vibrato. So we need lots of peaks (20-30 seems like the best, from their results) and the peaks need to not match any *particular* note's harmonics so all notes are similarly affected. Finally, the peaks need to be spaced enough that's there's roughly a -12dB trough between them. Much more than than that, and you get a "hole" and dead notes, less than that and you don't get enough movement in the notes.

So there are really two basic types of resonator bank you could build. One would be basically a sophisticated parametric EQ with several bands (say, 3, 4 or 5) with variable frequency, Q, and gain. The other would be a much larger bank of fixed filters. This would have individually variable gain, and could potentially have overall variable Q by wrapping feedback round the whole lot, as described in Bernie's Electronotes posting.

The first style is probably good for woodwinds, brass, and vocal sounds. The second is what you need for strings, at least according to that paper.

Ok, that's the results of my researches on the subject, and my understanding of what's been discussed here so far! I hope it helps.

Tom

==================
       Electric Druid
Synth & Stompbox DIY
==================




On 3 Jul 2017, at 16:43, ezion <ezion67 at gmail.com> wrote:

> Think the Elektor Formant resonator module might also be of interest to you. 
> Sorry no link, but the Formant schematics (a DIY modular from the late '70) are on the net for sure.
> Basically its a 3 band boost only parametric EQ.
> Personally now days when working with VSTs I just abuse a EQ plug-in.
> 
> Groetjes Theo
> 
> On Thu, Jun 29, 2017 at 12:43 PM, Elaine Klopke <functionofform at gmail.com> wrote:
> Hey list,
> 
> So I'm sure that I'm not aware of all of the variations, but....
> 
> I've run across the Scott Stites triple Wilson SVVCF (and still plan on building it!), and the Cwejman QMMF, and was thinking of cooking up my own. But how many filters is overkill? Above are examples of three and four filters in parallel with independent controls and a master section to alter them all together. What about five filters? Six? 40?
> 
> Just another silly question going through my brain..
> 
> Sent from my iPhone
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at synth-diy.org
> http://synth-diy.org/mailman/listinfo/synth-diy
> 
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at synth-diy.org
> http://synth-diy.org/mailman/listinfo/synth-diy