Applicability of the partial coherence function to the separation of noise sources in a semi‐anechoic field

Abstract

Two random noise sources for simulation and two wood product machines have been used to test noise source identification measurements. Noise signals are processed by two single channel FFT analyzers adequately syncronized and by a calculator with appropriate interfacings. The predicted value is compared to the actual value obtained in the absence of the second source. Comparison has been made between coherence due to incoherent noise sources with measurement contamination, and coherence due to partially coherent noise sources with no contamination. It is shown that artificial coherence of a certain level created by the measurement contamination causes more discrepancy on the predicted results than the same level of real coherence. It has also been shown that the smoothing techniques often used to compensate for average number cannot be employed when dealing either with separate noise sources with discrete frequency components or with a ground‐reflecting surface which creates an interference pattern. When narrow‐band analysis is used with an empirical correction formula the results show that even large and abrupt changes in the dynamic range and also ground effects can be predicted with good correlation.