Harry's Rant was: Guitar synths (Don's rant)

[WeAreAs1 at aol.com]

In a message dated 5/20/00 8:00:22 AM, you wrote:

<<MIDI is completely inappropriate for guitar for about a dozen
serious reasons.  It's also completely inappropriate for voice,
harmonica, sax, and a lot of other instruments for similiar reasons.
MIDI was designed for a small set of keyboard operations and little
else.  I think it's important to keep MIDI in perspective.>>

Even more to the point, here's something else to keep in perspective:  The 
guitar, simply by its own very nature, is already an almost infinitely more 
expressive instrument than the most expressive synthesizer out there.  Also, 
quite ironically, the acoustic guitar is, in many respects, more expressive 
than the electric guitar.  I'm not talking about the guitar's overall tonal 
possibilities - indeed, the synthesizer player has a much wider palette of 
unique sounds to choose from.  Rather, I am speaking about the guitar's 
ability to respond tonally to gestural input on the part of the player, 
within its naturally generated tonal voice.  Those of you who don't play 
guitar (or those whose don't play guitar very well) may not yet agree with or 
understand this, but those who have mastered some semblance of guitar 
technique will certainly agree.

That being said, here we have a situation where we've got a bunch of 
guitarists who wish their guitar synths could approach the expressiveness 
that they get from a normal guitar... Well, it's just not going to happen.  
Keyboard players don't even get that kind of expressiveness, baby!

BTW, in my opinion, the Roland GR-300 is (still) the king of all guitar 
synths, both in terms of response time and in gestural response (although it 
does have a fairly limited range of synthetic tone possibilities).  Unlike 
any of the others so far introduced, it even will respond (to a certain 
degree) to changes in pick position and angle.  However, I'd still like to 
figure out a way to get it to respond even better to gestural changes.  

One solution might be to run each string output into something like a bank of 
specially-tuned bandpass filters/detectors (sort of like a vocoder filter 
bank), then use the output of the detectors in some way(s) to control the 
synth's tone (not using the filters to filter audio, but using the detectors 
as control sources).  This detector bank would be used mainly to respond to 
attck gestures.  You'd probably also need to sample and hold the detector 
outputs at the time of each new note, to keep the detectors from trying 
follow the note's naturally decaying envelope.  Of course, you should 
probably ALSO have an envelope follower working in parallel with this system, 
but it wouldn't need to be frequency-selective like the attack gesture 
detector.  

This attack gesture detection system would run in parallel with the guitar 
synth's normal pitch-to-voltage (or pitch-to-MIDI) conversion system.  One 
might be able to experiment with this idea by using a regular vocoder, but to 
really work well, the bandpass filters would probably need to have their 
frequencies spaced more closely, with each filter having very narrow 
bandwidth.  Furthermore, the overall frequency range of the filter/detector 
bank would need to slide up and down with respect to the fundamental note 
that the player is currently playing.  This would guarantee reasonably 
uniform response up and down the fretboard.  You'd probably need somewhere 
around four to six octaves worth of filters to get it to reasonably respond 
to the typical range of guitarist gestures, all the way from gently playing 
with the "meat" of your bare fingers, to scraping the string with the edge of 
a very sharp guitar pick held at an oblique angle.  The filters would 
probably need to be spaced no more than 1/2 octave apart, and again, the 
frequency response of the entire bank would  need to slide up and down, 
relative to the current note.

Oh yeah - and you'd have to have a separate bank for each string.  Sounds 
like a nice weekend project, eh?

Of course, this still wouldn't come close to truly interpreting ALL of a 
guitarist's possible tone-influencing gestures, but it might be a start.

Michael Bacich

PS - As may already be apparent, I haven't really thought this 
filter/detector idea out completely.  For instance, I'm not quite sure how 
the individual filter/detector outputs (within one "string bank") would be 
combined for proper utilization.  Maybe some of you have some further 
thoughts on this?