Fan Noise Analysis Using Microphone Array

F3 - Active Noise Control & Noise Analysis

Fan noise source localization is currently a significantly investigated topic. Growing demand for comfort and various standardizations imposes to manufacturers to reduce the noise of their equipment or facilities. Unfortunately the fan is often one of the noisiest components. For a better understanding of the noise source generation, MicrodB has developed a localization algorithm for the analysis of rotating sources.

When considering the rotating sources, software makes the calculation grid rotate to be synchronous with the blades. Beamforming is done in temporal domain, for each sample, calculation point location is used to estimate the delay applied on the signal. It results in a temporal signal with irregular sampling frequency. Those signals are then interpolated on regular time scale.
Classical time domain beamforming would consist in a summation of the delayed signals from each microphone, but in MicrodB algorithm presented in this paper, they are swapped in frequency domain, used to build the cross spectral matrix, and the diagonal of this matrix is deleted.
Validations of this algorithm have been done with a tonal source. It has been successively characterized with and without rotation, which allows the localization and level estimation validation. This method is being tested for the analysis of an industrial high speed fan, at the ISAE laboratory.