I used a Sennheiser MD 421 dynamic microphone connected directly to a Soundblaster Live soundcard.  The soundcard converts the sound from analog to digital and stores it digitally.  The gain on the microphone was at the maximum for each sample.  The microphone was on a stand on a table, about 3.5 feet above the ground.  The players were standing about 9 feet away from the microphone for all samples, unless otherwise specified.  I took samples at two different distances, and decided that further away better represents what a listener would hear.  Each player played one note without vibrato. Each sample is approximately a one second piece of this note in order to analyze different parts of the same note.  I used a LabVIEW program to execute the Fast Fourier Transform and then plot the resulting power spectrum of each sample. Figure 2 shows the setup.

    I took samples on four different flutes with three different players.  I will refer to the players as 1, 2, and 3.  Table 2 shows the characteristics of each flute.  A closed hole flute has solid keys to cover the tones holes.  An open hole flute has holes in each key which must be covered when pressing them.  Foot refers to the foot joint of a flute, with a B foot longer than a C foot because it has one extra key.  The lowest note that a flute with a B foot can play is a B (247 Hz).  Otherwise, the lowest note on a flute is the C above that. Category refers to the manufacturing quality of a flute.
 
 

                                                      Table 2 - Flutes used in data collection

 The LabVIEW program that plotted the power spectra took 215 samples at a rate of 44,100 samples/sec. Therefore, the measurement uncertainty of the program is +/- 1.35 Hz.  For a power spectrum of A at 440 Hz, the worst uncertainty would be 0.3% of the frequency measurement.  The measurement uncertainty of the soundcard is +/- 0.05 dB.  However, I did not investigate the actual sound pressure level at the microphone or the output sound pressure level of the flute.  The sound pressure levels are not needed because this project is a comparative analysis, and I normalized every plot by the amplitude of the fundamental frequency.  The experiment is repeatable in terms of the frequency values of the power spectrum peaks, however, the amplitudes of the peaks are different.  The values of the peaks vary by flutist and due to flute distance from the microphone and volume of the note.  The 95% confidence interval for one sample was found to be as much as +/-30% for each spike in the power spectrum.  The confidence varies greatly with each sample due to the large human factor uncertainty.

Abstract
Introduction
Experimental Setup
Results
Discussion
Conclusions
References