The research undertaken aims to the development of a new numerical predictive technique, capable of producing such a detailed simulation, so that it can be used for the creation of “virtual” sound samples, theoretically indistinguishable from binaural recordings inside a real car. This way, a proposed (not existing yet) sound system could be compared with the existing ones by direct listening tests. This paper describes the implementation of the numerical model, based on Boundary Elements for low frequencies and Pyramid Tracing for higher frequencies. The results of the two models are merged together, producing a binaural impulse response which can be convolved with the music signal. The performances of the system were evaluated by listening tests, in comparison with experimental measurements taken in a 1:5 scale model of a car compartment.

Download

In this paper a non-linear loudspeaker model, which accurately reproduces the low frequency behavior, is presented. This description, derived from an extension of the well known Small-Thiele equations, requires far less computational time and memory space than generic non linear structures. Moreover a noticeable further reduction of the number of operations and of the memory cells required has been achieved by means of a multirate architecture. Inversion of the proposed model allows digital prefiltering of the electrical signal in order to compensate for the nonidealities of the electro-acoustic conversion. The above filter structure implemented on a digital signal processor, placed between the audio signal source and the power amplifier allows effective compensation of loudspeaker linear (both magnitude and phase) and non-linear distortion. Measurement results obtained with a commercial woofer are discussed

Download