A Comparison of Seismic Array Processing Methods

This paper discusses a number of alternative schemes for combining the outputs of the seismometers in an array so as to maximize the improvement in signal-to-noise ratio while imposing a minimal distortion on the waveform of the signal arrival. For gaussian noise, these requirements lead to linear processors. A brief review is given of the class of such processors that involves a single delay or a single delay and amplitude weight in each sensor output before summation. The more general class of linear processing involving a linear filtering operation on each output before summation is then described. The two principal procedures of this kind are maximum-likelihood and Wiener processing. The first of these results from imposing the condition of zero frequency distortion of the signal; the latter results in an optimum tradeoff between noise and signal distortion in the output. Experimental results are presented showing the result of using these various schemes on small numbers of seismometers at installations in Arizona (TFO) and Montana (LASA).