No abstract

Long enclosures are spaces with nondiffuse sound fields, for which the classical theory of acoustics is not appropriate. Thus, the modeling of the sound field in a long enclosure is very different from the prediction of the behavior of sound in a diffuse space. Ray-tracing computer models have been developed for the prediction of the sound field in long enclosures, with particular reference to spaces such as underground stations which are generally long spaces of rectangular or curved cross section. This paper describes the development of a model for use in underground stations of rectangular cross section. The model predicts the sound-pressure level, early decay time, clarity index, and definition at receiver points along the enclosure. The model also calculates the value of the speech transmission index at individual points. Measurements of all parameters have been made in a station of rectangular cross section, and compared with the predicted values. The predictions of all parameters show good agreement with measurements at all frequencies, particularly in the far field of the sound source, and the trends in the behavior of the parameters along the enclosure have been correctly predicted.

This paper describes two methods of modelling a curved surface in ray tracing computer models: an approximate method in which the surface is modelled as a series of planes, and an exact method which uses the precise analytical description of the surface. The focusing effects predicted using the two models in a hypothetical cylindrical space have been examined and compared. It was found that the approximate method of modelling the curved surface predicted focusing effects that were significantly different from those predicted using the exact method. However, the larger the number of planes used in the approximation, the closer the agreement between the two methods. The effects of source position and number of sources on focusing were also examined, and showed that focusing effects are stronger when the source position is on the axis of the cylinder. Propagation of sound across and along the cylinder has also been studied.

A ray tracing computer model has been developed for the prediction of the sound field and speech intelligibility in long enclosures ofrectangularcross section. The model has beeen validated using measurements made in a station on the Hong Kong MTR. There was close agreement between measured and predicted values of SPL, EDT, C50, D50 and the speech transmission index.