Acoustic imaging using nearfield holography

Nearfield acoustic holography (NAH) is a relatively new technique that can be used for non-contact imaging of acoustic sources. One of the practical limitations of NAH is the extremely large number of sample points needed for accurate imaging. With present technologies, the only feasible way of incorporating so many sample points is by improving the resolution of the hologram plane data by sampling with a sparsely populated array of transducers, which can be moved to collect several sets of samples. This method of sampling is known as `translation´. For any form of translation to be effective, two criteria must be met. Firstly, the waveform emanating from the source must be repetitive in nature over a relatively short period of time. This obviously excludes the use of translation with sources that are of a transient nature. Secondly, the coherence of all of the samples in the array must be maintained, that is, the sampling after a translation must be started at some constant interval of time with respect to the start of the period of the source. Two numerical methods are described in this paper that can transform incoherent sample sets into coherent data. This now means that `free running´ sources can be imaged without the need for a synchronisation pulse from the source to trigger the sampling. Another problem that is common to virtually all real sampling systems that work in the frequency domain is a phenomenon known as `spectral leakage´. This effect arises when the period of the sample window used by the discrete Fourier transform (DFT) is not an exact multiple of the period of the source. The effects of spectral leakage are that the shape of the spectrum becomes distorted, which in turn reduces the true magnitude of the local peak frequency, and increases the magnitude of the side lobes. This paper describes a new algorithm that modifies the sample data to minimise spectral leakage, as well as maintain the true magnitude of the frequency components. In addition, a much higher resolution of frequency extraction than is normally associated with DFT´s is now possible. Numerical examples from a variety of different sources are given