Comparison of Different Analysis Techniques to Decompose a Broad-band Ducted Sound Field in its Mode Constituents

In this paper two different experimental techniques for decomposition of broadband ducted sound fields into its mode constituents are compared in a bench test. Broadband mode analysis method I that was published in a previous paper 1 makes use of the spectral crosscorrelations of one reference sensor with all other sensors. The new broadband mode analysis method II takes into account the spectral cross correlations of all sensors in combination with a beamforming approach and subsequent decorrelation step for modes featuring the same azimuthal order. For comparison, both techniques were applied to the same fundamental experiment. Broadband sound fields of different signatures were generated in a laboratory no-flow rig by use of up to 16 wall-flush mounted loudspeakers that were fed with independent random signals from a digital signal generator. The cylindrical flow duct of diameter 50 cm and axial length of 400 cm was equipped on both ends with non-reflecting terminations. The resulting sound field was acquired by means of a wall-flush mounted microphone array with 64 microphones. The microphone signals were acquired simultaneously over 3 minutes, allowing for a statistically accurate crosscorrelation data analysis. Both broadband mode analysis techniques I and II were used to compute the mode amplitudes and the resulting induct sound power for each narrow-band frequency component. The techniques were benchmarked with regard to their ability to distinguish coherent and incoherent modes.