similarities to those used in traditional CFD techniques, the nature of the problems addressed and the results de-This paper describes the application of a three-dimensional computational aeroacoustics (CAA) methodology to the prediction of sired are very different in the case of CAA studies. Some jet noise. The technique has been implemented using parallel comof the critical issues that are typical of CAA are: the nondisputers. In this approach the nonlinear disturbance equations are sipative and nondispersive character of acoustic waves; the solved in a conservative form using a finite-difference based techextremely low amplitude of acoustic perturbations comnique. A fourth-order optimized dispersion relation preserving pared to the mean flow values; and the high frequencies scheme is used for spatial discretization and a fourth-order classical Runge-Kutta scheme is employed for temporal discretization. The of waves that need to be resolved. Hence special attention three-dimensional CAA code has been parallelized using a domain has to be given in the development of numerical techniques decomposition strategy in the streamwise direction. The calculato address these issues, particularly the minimization of tions are carried out on both IBM-SP2 and SGI Power-Challenge dissipative and dispersive errors. The following section parallel computers using message passing interface routines to gives a brief review of current CAA methods used in jet facilitate exchange of boundary data between adjacent nodes (processors). Excellent parallel performance has been obtained using noise predictions. the present code. Acoustic results are presented for a perfectly Hixon et al. [5] used a solution of the linearized Euler expanded supersonic axisymmetric jet under harmonic and random equations to predict the jet noise characteristics of an axiinlet conditions. Results are given for both the instantaneous and symmetric supersonic jet. Analytic mean flow profiles were averaged flow and acoustic variables. Comparisons are made bespecified in their calculations. A large eddy simulation tween the predictions and experimental data.