<p>Extensive and precise measurements have been made of instantaneous sound pressure distributions concerning 44 representative orchestra music which were composed from the eighteenth to the twentieth century and frequently have been performed. Amplitude distribution F(x) of a signal X(t) is a probabilistic expression for the signal and is defined by F(x)=T_x/T where T_x=Σt_i(X&ge;x) and T is measurement time. (Fig. 1. ) The measuring equipment of amplitude distribution, of which the diagram and the performance are shown in Fig. 2 and Table 1 respectively , is designed and constructed so as to suffice the requirements derived from the satistical relations (2) and (3) to the error caused by sampling. Some preliminary experiments were conducted, to approve that the distribution of orchestra sounds has the property feasible to be measured without a considerable amount of degradation even through the conventional measuring system including microphones and tape recorders whose phase distortion is usually not much minded, and then to know about the linearity characteristic of the tape recorder. It is first found that the distribution of orchestra sounds is scarcely deteriorated by the phase distortion and is almost symmetric for either sense of the signal. Figs. 3 and 9 indicate these properties respectively. In the prestent report, although invariability with the phase distortion of the tape recorder only is presented, it is convinced, with the results of other experiments, that hardly any error is feared to be induced by other types of phase distortions. Besides these, referring to the linearity of the tape recorder shown in Fig. 4 in the case with a test material which includes wide frequency components, the maximum VU meter indication is obtained below the level of which the reproduced orchestra signal is not affected by the non-linearity to reduce the level of picked up sound. Along with these preparatory investigations, much concern is also paid in minimizing the amplitude frequency distortion. Fig. 5 - 7 show the frequency responses and tolerances of microphones and the tape recorder used. In picking up orchestra sounds for measuring material, the same unidirectional condenser microphone for broadcasting practice placed in a typical position was employed. The recording level was set, according to the maximum VU meter indication during rehearsal, at the subtracted level from the overload point of the tape recorder by several decibels to record the signal without conceivable distortion, considering the probable deviation from the peak sound pressure in rehearsal. The mesurement of instantaneous amplitude distribution was conducted only for the half-wave signal. The sound pressure level is assessed by the calibration of the microphone sensitivity. Some examples of the measured distributions are shown in Fig. 10. They are almost exponentially shaped over much of the range except in the portion at the lower level. Therefore, it seems one of the most natural ways to approximate the obtained data by the normalized composite distribution (4) consisting of three terms, the first two of which represent the exponential distributions for the high-level and low-level signals respectively, and the last, δ distribution represents such extremely low-level signals as attenuated reverberation sounds. Thus, the measured distributions are processed into the normalized distributions, each set of coefficients of which represents each form of the distribution for a particular sound of performance. As to information on absolute sound pressure, on the other hand, it is represented by both the rootmean-square pressure V and the peak-factor PF which is interpreted here as the difference between the peak and the r. m. s. level. Adding to these kinds of coefficients, the peak level in S. P. L and Trms, the percentage of time for which the signal exceeds the r. m. s. pressure are obtained. These</p><p>(View PDF for the rest of the abstract.)</p>