Measured reproducibility of clarinet spectra

Abstract

Published spectra show little consistency beyond a tendency toward strong odd partials. Neglect of radiation pattern and wave‐statistics fluctuations in rooms, join with inadequate specification of the player's task to cause this variability. We have studied 0.5 s alternating C4, D4 tones lasting about 35‐s total at mezzoforte level (reed just beginning to beat), played on six clarinets, A, B♭ (three), C, E♭ All instruments were optimized to similar criteria over the past decade by one of us (AHB) who also made each its own mouthpiece (the bores were dissimilar). The instruments all have been used by major players. The player (AHB) walked about while playing (0.75 m/s), as did the two persons carrying the recording microphones, in a room V = 8000m3, T60 = 1.2 s. Source and mike position changes, plus mode perturbation effects from three people, assure the equivalent of at least 75 statistically independent data points for each component of each tone. The SPL for the first five harmonics averaged over all twelve tones are 50, 22.6, 48.8, 27.0, 48.3 dB (σ = 2 dB). Room fluctuations are within ±1 dB (95% confidence). Most of the remaining variation comes from interinstrument differences and not player/instrument instabilities. These results are consistent with a preliminary experiment analyzed by Karla Brasaemle. [Work assisted by NSF.]