> I agree. So one pulse difference (the smallest difference that can be > read) would be 1/210 of a cent, not 1/6000 of a cent, as you claimed. > even that requires that you be able to trigger your counter to an > uncertainty of no more than 100 nanoseconds, which is hard to believe for > a voltage that takes over 36,000,000 nanoseconds to rise through the > trigger threshold. No way, man! In fact, if you derive your measurement > from the period of a single cycle of 27.5 Hz and don't average over > multiple periods, your accuracy is probably no better than 2 full cents at > A0. Partials can be picked up by even the crappiest pickup, and these > partials will distort the sine wave of the fundamental so that the > one-shot triggering time has an uncertainly of many many clock pulses. > > Robert Scott > Real-Time Specialties Read the whole post. The 1/200th of a cent was only for the $6 Radio Shack version. Higher frequency oscillators and faster counters are used in the production model...not because it was necessary, but because it's what was available and was used for other reasons. As I also mentioned, this is still much more accuracy than is necessary. The big benefit of this method is that it can be done quickly, in one cycle. Why do you have to be so abusive and declare that an already built and tested device is impossible? The signal is quite clean. Part of this is due to the pickup's inability to detect partials (it's not a microphone, it reads vibration at the spot where you place it), but much of it happens in the sustaining loop. If you take the signal from the pickup, amplify it, then filter it and apply it back to the string, you not only filter the tapped signal, you also *drive* the string at its fundamental. The result is a fundamental sine wave that is not visually discernable from a perfect sine wave on the oscilloscope. When it is squared, the result is a beautiful, steady square wave. As I mentioned in another post, I test it by sustaining a string and observing the actual counter numbers produced, refreshing every half a second or so. This appears as a number that keeps updating over and over. What I find is that it does indeed fluctuate slightly. For A0, for example, it can vary by up to six or seven counts. There could be several reasons for this including natural fluctuations in the string frequency, or variations in reading between different waves. The fact of the matter is: I don't care why. These variations in the period number are still well within reasonable accuracy. To average them might gain me some more accuracy, but, again, I don't care. It's already more accurate than necessary. Don A. Gilmore Mechanical Engineer Kansas City
This PTG archive page provided courtesy of Moy Piano Service, LLC