The development of quieter aircraft engines is only possible if the sources of sound in the engine are known. They need to be localized, identified and diagnosed in as much detail as possible. Acoustic measurements with single microphones, which are routinely performed during the acoustic certification process of aircraft, can only provide limited information. They can be used to analyze the spectral content and the overall power, but they can provide only indirect information on the sound sources themselves. Only when the location of a sound source is known, the source mechanism can be analyzed and a thorough understanding of the source mechanism is a prerequisite for the development of a noise abatement device or procedure. Microphone arrays, or phased arrays, can be used to determine the position of sound sources on stationary or moving objects. Localising sound source with microphone arrays is an established technique that is widely applied in research and industry. The DLR department of engine acoustics has been using microphone arrays for a long time and has contributed to the development of the technique. First applications in the department date back to 1987, when high speed trains were analyzed with linear microphone arrays. The technique has been refined and extended to the analysis of aircraft flying over large two-dimensional microphone arrays.
DLR, TU-Berlin and the Gesellschaft für angewandte Informatik (GfaI) organize the biannual Berlin Beamforming Conference where specialists present and discuss the latest developments in the field of sound source localisation and analysis with microphone arrays. More information can be found on the website www.bebec.eu
Equipment
The DLR department of engine acoustics owns several data acquisition systems with up to 256 input channels. The data can be sampled simultaneously with a resolution of 24 bit and sampling rates of up to 50 kHz. Extra channels are reserved for auxiliary signals, e.g. time codes or trigger signals for engine shaft speeds. Several types of microphones are available for different tasks: precision condenser microphones with 6.35 mm membranes or robust electret microphones for outdoor applications. Depending on the desired precision and measuring environment, a suitable combination of equipment can be set-up for a specific measurement task.
Classical beamforming and equivalent source methods
The basic principle of sound source localisation with microphone arrays is relatively simple: the signals of many microphones that are distributed over a wider area are recorded and delayed relative to each other in order to compensate the propagation times between a particular source position and every microphone in the array. When the delayed signals are summed up, contributions coming from the source position add up constructively because they are in phase. Contributions from sources away from the focus position are not in phase and they interfere destructively. This method is called sum-and-delay beamforming. However, this simple approach does not provide quantitatively correct information on the sound sources because the geometric properties of the microphone set-up, the distance and angular direction to the source all play a role. Also, the averaged signal is not well suited to investigate source directivities. Modern methods for the analysis of microphone arrays, e.g. DAMAS or functional beamforming compensate the imaging properties of a microphone array and provide quantitative results.
SODIX
Break-down of the sound field of a full-sized aircraft engine into the contributions from individual engine components calculated with SODIX
For the analysis of the source directivity of aircraft engine noise, the DLR department of engine acoustics developed an inverse array method called Source Directivity Modeling in the Cross-Spectral Matrix, or SODIX in short. This method models the absolute contributions of a distribution of directive sound sources by fitting a set of equivalent sound sources that provides the best model for the cross-spectral matrix measured with the microphone array. The algorithm minimizes the squared difference between the measured cross-spectral matrices and the cross-spectral matrix calculated from the equivalent source distribution. The source amplitude of every equivalent source is determined in the direction of every microphone – this provides the directivity of the model point sources. Measurements of full-size engines on static test-stands have been analyzed with SODIX. The results show that SODIX is able to model the broadband sound sources of the engine with a very fine spatial resolution. The far-field directivity of the sound field generated by the engine can be reconstructed by propagating the equivalent sources into the far-field of the engine. Another application of SODIX is the localisation of noise sources in wind-tunnel measurements, where the individual microphone signals are distorted by uncorrelated aerodynamic noise. In such a situation, SODIX can be applied to the cross spectral matrix after removal of its main diagonal, which contains the uncorrelated noise. In contrast to classical beamforming methods, SODIX can cope with the diagonal removal in a mathematically correct way and reduce the influence of uncorrelated noise.