The detection of intermittent near-field noise sources is achieved by continuous wavelet transform of far-field acoustic data. The similarities between signals from three far-field microphones in the annular cone of acoustic radiation from coherent sources in a Ma = 0.6 cold jet are used to identify some of the loudest events. Cross-correlation of relevant band-passed excerpts identifies the most reliable matches, which occur at the rate of one every 20 dimensionless time units on average, and are intermittent in both time and frequency with no obvious pattern. Processing our database produced a catalog of over 9500 individual events with quantitative properties including the magnitude, frequency and time of detection at each microphone. The differences in detection times (lags) are distributed smoothly around the peak of cross-correlations for the raw data. We show how source magnitude and frequency depends on lags, and relate the lags to their relative source location.
展开▼
