[sdiy] guitar pitch detection, was: Mighty quiet lately...

[John Richetta]

On Oct 3, 2010, at 5:32 PM, Paul Perry wrote:
> The problem is not detecting the pitch - it is detecting the pitch  
> QUICKLY, and by quickly I mean within less than a full cycle.

Ah, fair enough summary, and I guess did already knew that, having  
played with some gear that does this.  But even the better gear I've  
played seems to still be somewhat glitchy - VG88, VF-1, GR-1, etc.   
So, perhaps there is still some room for improving pitch detection  
reliability?

An unstated assumption in what I wrote was that I wasn't expecting  
miraculously fast pitch detection - I presume any DIY analog circuit  
will be far from ideally fast, and will probably need to be used in  
combination with relatively direct signal, or else played in a studio  
type situation (where delay can perhaps be better tolerated and  
compensated for).

> Can it be done?
> Yes - if you track the exact position of each string, and use
> this information to drive a DSP modelling algorithm. I believe this  
> approach has been used successfully.

Sure; now that you mention it, I do recall hearing that about the  
VG-88 (and kin).

Note that, strictly speaking, detecting wave slope even for less than  
a full cycle (as is necessary primarily for low notes) is not  
*necessarily* different than measuring the output response of some  
filter.  That is, a tuned filter responding to this input might  
correctly identify it as being in the right range pretty quickly  
(though I must admit that the control would need to be smarter than I  
describe to do that).  Obviously, practical comb filters do incur  
especially egregious delay, and generally speaking, using the  
knowledge of how waves propagate down a string, especially initially,  
is useful for fast response.

So, yeah, what I described is probably junk - it certainly doesn't use  
that info.  Sorry for wasting the bandwidth.

-jar