Polished vs Un-polished Magnets

[Joe Gwinn]

RE: cepstrum and other methods for pitch detection:
Now that I've researched it, it seems like all pitch detection methods suck, varying only in degree.
Reported latencies are nominally ~50-60ms latency, best case 15ms.

Depends on SNR, I'll bet. And the pitch period. I guess my question is sucks compared to what?

Human hearing seems pretty good at deducing pitch given only the upper harmonics. But then again, humans have delays in the tens of milliseconds as well.

[salvarsan]

Depends on SNR, I'll bet. And the pitch period. I guess my question is sucks compared to what?

Wired frets. Laser doppler.

Re: Human pitch perception latencies
Anyone of any proficiency on guitar has trained right and left hand technique without the conversion latencies added in. The 50ms period is in same order of muscle "stay-alive" micro tremors and the eye flicker fusion processing rate. This latency was also significant in Allan Holdsworth's choosing a Synthaxe instead of a guitar outfitted with a MIDI pickup -- it tracked his legato left-hand technique without delay.

Right now, the one to beat is Fishman's TriplePlay, a wireless conversion device for electric guitars intended to control synthesizers. Fishman tries very hard not to say MIDI in the ad copy and there is no mention of conversion latencies, but the $400 integrated hardware+software package generates favorable reviews.

[Mike Sulzer]

Depends on SNR, I'll bet. And the pitch period. I guess my question is sucks compared to what?

Human hearing seems pretty good at deducing pitch given only the upper harmonics. But then again, humans have delays in the tens of milliseconds as well.

From a signal processing point of view, higher harmonics are what you want to use, as high as SNR allows. This gives you the most periods in a given time for the frequency measurement. To get the fundamental, you need the spacing between harmonics just accurately enough to avoid ambiguity.

[David Schwab]

Right now, the one to beat is Fishman's TriplePlay, a wireless conversion device for electric guitars intended to control synthesizers. Fishman tries very hard not to say MIDI in the ad copy and there is no mention of conversion latencies, but the $400 integrated hardware+software package generates favorable reviews.

It is MIDI. You need a computer to run it. I don't know if it's just the video, but I can see the latency. He says it tracks well though.

Here's a demo of it:

Demo: Fishman TriplePlay Wireless MIDI Guitar System

(there is written discussions about it at the link above)

[Mike Sulzer]

From a signal processing point of view, higher harmonics are what you want to use, as high as SNR allows. This gives you the most periods in a given time for the frequency measurement. To get the fundamental, you need the spacing between harmonics just accurately enough to avoid ambiguity.

This requires some explanation; have the time now.

Suppose that you sample your signal for time T. The basic frequency resolution of a Fourier analysis is 1/T. The error in frequency determination can be some fraction of that with good signal to noise ratio. This applies to whatever is there, be it fundamental or harmonic.

Suppose you measure the fundamental, and you have an error fe. Suppose you measure a harmonic and the error is also fe, implying that you have good SNR on the harmonic. If you know the harmonic number, you divde by it to get the fundamental. The error divides down, too. That is the good part. But you do not know the harmonic number, and so you measure several harmonics, enough to see what the spacing between harmonics is (at least two, but maybe more to be sure). Once you know the spacing, you know the harmonic numbers by dividing, and then each harmonic gives you a measurement of the fundamental. If you are limited by random noise, the measurements are independent, you can average them all to get an improved one. The result is a much better measurement than using the fundamental directly. This means that you can reach a desired level of accuracy in a shorter time.

[Joe Gwinn]

Wired frets. Laser doppler.

Re: Human pitch perception latencies
Anyone of any proficiency on guitar has trained right and left hand technique without the conversion latencies added in. The 50ms period is in same order of muscle "stay-alive" micro tremors and the eye flicker fusion processing rate. This latency was also significant in Allan Holdsworth's choosing a Synthaxe instead of a guitar outfitted with a MIDI pickup -- it tracked his legato left-hand technique without delay.

It's true that ~50 ms is the closed-loop response time of college sophmores in the classic light and telegraph key experiment: Watching the light intently, click the key as soon as the light turns on. Hungry rats are faster, if I recall.

But it isn't true that people cannot do things faster, after some practice. The hand moves become fluid gestures. A while ago I was watching one guitarist with spider fingers.

You literally could not follow his fingers - they blurred. The hand is quicker than the eye.

[Mike Sulzer]

It's true that ~50 ms is the closed-loop response time of college sophmores in the classic light and telegraph key experiment: Watching the light intently, click the key as soon as the light turns on. Hungry rats are faster, if I recall.

But it isn't true that people cannot do things faster, after some practice. The hand moves become fluid gestures. A while ago I was watching one guitarist with spider fingers.

You literally could not follow his fingers - they blurred. The hand is quicker than the eye.

There is a big different between sense-analyze-respond and executing a planned set of actions. That is why the guitar playing can be so fast!