Shaping Guitar Tone to Narrow Pulses??

[Jaicen_solo]

Thanks for that Dan, saved me some hassle there! Still not prototyped yet, i'm still working on a Rocktave for my friend, so it's ok.

Gez, i've spent some time reading some of your other threads, particularly with reference to the AMS-100 pluck detector, and I understand how the schem you posted works now, seems like a very elegant solution, one I would actually define as simple. However,  I was wondering if you knew of a way to treat the guitar signal, and extract a pulse at every zero crossing?
My theory is that a 4016 could be used to discharge a cap charged through a constant current source, when a signal drops below 0v. If I was working off a bipolar supply, say +/-12v, the waveform will be positive for 50% of the period, and negative for the remaining period. So the cap would charge when the signal is above 0V and discharge when it's -'ve.
My theory is that this will improve tracking, since it will eliminate false triggering at different levels of the waveform (ie, half way up the slope etc), andI guess will track the lowest fundamental and ignore harmonics. It would also make it much easier to define a trigger threshold. Actually, this isn't my idea, I stole it from Roland but I don't think they'll mind
So, ermm yeah. Any thoughts?

[gez]

Jaicen, I'm afraid you've lost me.  Do you mean a fundamental extractor?  The problem with real world guitar signals (and I'm talking monophonic here) is that if you 'extract a pulse at every zero crossing' you don't always end up with the fundamental.

If you have a scope, run your guitar through it and you'll see what I mean.  Often, due to harmonic distortion, the waveform will momentarily dip back bellow/above the 'zero crossing' point as it passes through it, resulting in false triggering - you often end up with frequency doubling.

I recently came up with a convoluted fundamental extractor.  It came about after running my guitar through the scope.  I observed what was happening and thought 'hmmm, I know how to deal with that!'.  What I didn't bank on is that switching to a different pickup (I only checked the neck one) resulted in a different type of distortion of the signal, and the circuit I designed actually made tracking worse when the bridge PU was used.  Humble pie all round!  I figured out a fix, but it's remained on the breadboard untested for the past couple of months... no idea if it works.

The idea (and I'm not claiming it as an original) was to sample only the peaks of a waveform and use them as a reference for settin the thresholds of a Schmidt type squarer.  The advantage of doing this is that you avoid all the crud during the zerocrossing.  Flipflops and some logic helped knock the signal back in phase.  Complicated circuit, though the one on my breadboard now is greatly simplified.

[gez]

Incidentally, that f-tov converter works 'backwards' to how most people would want: voltage decreases with frequency.  It doesn't matter for a frequency doubler, but if you wanted to use the sawtooth created by the circuit you'd need an OTA fed from an increasing voltage (with fequency) to keep amplitude steady.  In which case, you need to reference everything (caps and switches) to the positive voltage the constant current source is run off (op-amps still need to be powered from a higher voltage).

[Jaicen_solo]

Hmm, I knew there'd be a reason nobody had DIY'd it already!
I figured the zero crossing thing would be the a fundamental extractor by default, since it would ignore peaks-on-peaks and only be triggered by a transition from +'ve to -'ve. As you've probably ascertained, I'm not all that knowledgeable about how to design this, but I do have a lot of time to sit and think (at work  ).
Maybe if I was explicit about what i'm planning that would help. The idea is to to take a heavily filtered signal from a hex pickup like a GK-2a, and process each of the strings to extract as close as possible a fundamental pulse wave. I'd then use this to drive the clock input of a CEM3396, or to directly generate a synth wave (which may be the best option).
In the latter case, your initial design would probably work quite well, as I could then feed the output to a schmitt trigger to square things up, then on to a 4017 to output intervals down. That would give me the option of +1oct Saw/Sq, down to -2octaves, which would be pretty cool. However, that option requires a seperate VCA + VCF for each string, whereas the CEM3396 simplifies things a great deal, and can probably deal with some jitter anyway.
All of this is a bit pie in the sky really till I get something breadboarded and see how the 3396 responds. Or until Paul Perry relents and sells me his GR300