Abstract
The well-known equivalent sound level (L_<Aeq>, abbr. , L_<eq>) adopted for a unified acoustic description of environmental noise by ISO 1996 has now gained some widespread acceptance, for example, as a measure of sound exposure. Nevertheless, in a theoretical study of L_<eq> statistics, considerable efforts must be yet spent for establishing the effective measurement procedure of L_<eq>, especially from the methodological point of view. This paper discusses the effect of the sampling period, Δ, and the sample number, N, on the L_<eq> statistics in a constant measurement time, T(=NΔ). The theoretical probability expression of general series expansion type for the average energy, E(L_<eq>=10log E/E_0, E_0=10^<-12>W/m^2), proposed here has a gamma distribution as its first expansion term and also reflects various types of linear and/or nonlinear (multi-dimensional) correlation properties for the sound power fluctuation wave into its each expansion terms. Theoretical and experimental results demonstrate the fact that a fluctuation probability distribution form of L_<eq> converges to a certain proper probability form with non-zero variance, even when N tends to infinity as Δ approaches smaller than a Nyquist period (1/2W, W: equivalent frequency bandwidth of sound power fluctuation) under the condition of a constant T. The effectiveness of the proposed theory has been confirmed experimentally too by applying it to the actual traffic noise data observed at the central area of large city.
