Uncertainty associated with in-situ frequency-response estimation by reference-sensor comparison at infrasound monitoring sites

Abstract

In-situ measurement of the frequency-response of infrasound array elements has proven to be a useful tool in the assessment of element performance. In order to transition to a true calibration process, the uncertainties inherent in the method must be determined. It is critically important to distinguish between bias errors and random errors and to recognize that the ambient pressure fluctuations are typically not stationary in a statistical sense. The time evolution of the cross-spectrum is particularly useful for identifying non-stationary behavior and for isolating high-quality data intervals. Three important cases are tractable: high coherence between the reference sensor and the infrasound element; low-to-moderate coherence resulting from uncorrelated noise in one channel; and moderate coherence resulting from uncorrelated noise in both channels. For a fixed number of averages, the confidence limits for the frequency-response estimate are often considerably tighter than the corresponding limits for the estimated spectral densities.