Sensor Testing Timeline

The performance testing at BSL occurred in stages, with most work being performed between July 2006 and January 2007. The calibration of the data logger took over three weeks due to a number of back-to-back system failures, including the power supply, and dead data channels. During three weeks in August 2006, the initial six STS-1's (four horizontal, two vertical) were set-up as near identical (baseplates, cables, bell jars etc.) as possible.

It was during this portion of the testing that sensor to sensor coherency issues arose. The vertical sensors showed high levels of coherency, while the slightest misalignment of the horizontal instruments caused unacceptable incoherence in the horizontal instruments. All possible causes for the incoherence were investigated and documented during the next 3-4 weeks. Possible causes included system timing, grounding, vacuum levels, and the seismometers themselves. Over a period of a week, the horizontal instruments were rotated slightly until the coherence began to improve in a predictable fashion.

Thus, seven weeks after BSL began to work on this project, the first meaningful tests of the Metrozet electronics began. Each set of electronics was paired with each of the seismometers in succession, with the pairing lasting at least twenty four hours. Approximately three weeks after this initial rotation through the seismometers was completed, Metrozet produced the second set of prototype electronics. Progress continued in similar two to three week increments until January of 2007, at which time the new electronics and calibration systems had been paired with and evaluated on seven different STS1 seismometers.

Figure 3.27: Results for two vertical component STS-1's (HHZ and HH2) and two horizontal component STS-1's (HHE and HH1) in the presence of a large seismic signal. The event is a Me 8.1 earthquake which occurred 87.9 degrees WSW of Berkeley at 20:39 UT on 2007/04/01. Shown are the signal PSD (red), the noise PSD (blue) and the coherence (brown) for each sensor. In all tests, the corresponding BKS STS-1's are used as the reference signals in the analysis. In the presence of large seismic signals, the coherence is typically close to unity at all frequencies below the 5 Hz high-frequency corner of the BKS reference STS-1's. Note the relatively high noise PSD level on the horizontal components in the vicinity of 0.1 Hz. This is due to a slight misalignment of the sensitive axes of the horizontal components. Several time consuming trial and error iterations in aligning the horizontal components are required to lower the horizontal component noise PSD. Three continuous hours of 200 Hz data are used by the scn_psd algorithm. scn_psd parses the data into 32 non-overlapping samples, applies a hanning window, corrects for the effects of the hanning window, scales the data to ground motion, calculates the FFT, and stores the resulting complex spectral values for each sample. At each frequency, the RMS signal PSD is calculated from the average of the complex spectral values, coherence is calculated from the averaged complex spectral cross product, and the RMS noise PSD is then determined from the product of the signal PSD and (1 - coherence). The method is described in detail in Gardner (1992).
\begin{figure}\begin{center}
\epsfig{file=rau07_1_1.eps, bb=100 180 535 592, clip=, width=8cm}\end{center}\end{figure}

Figure 3.28: Results for the STS-1's in the presence of background noise. The traces are the same as in Figure 3.27. The lower and upper frequencies at which the coherence degrades from near unity varies among the sensors. Coherence bandwidth is a measure of the performance of the sensors.
\begin{figure}\begin{center}
\epsfig{file=rau07_1_2.eps, bb=100 180 535 592, clip=, width=8cm}\end{center}\end{figure}

Berkeley Seismological Laboratory
215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760
Questions or comments? Send e-mail: www@seismo.berkeley.edu
© 2007, The Regents of the University of California