sensors buried cables. 5-10 m. ~100 m. Electrodes magnetic sensors other site equipment telemetry ... It is cheap, easy to make, and .... Note that in each time-domain plot, the DC shift is removed and the data are nor- malized to a range of ...
NH31A-1338
Stanford – USGS Ultra -Low Frequency Electromagnetic Network: Hardware Developments in Magnetometer Calibration and Data Recording Daniel Bowden *, Henry Engelland-Gay , Aaron Enright , Jarrett Gardner , 1 2 2 Simon L. Klemperer , Darcy K. McPhee , Jonathan M. Glen 1
1
2
3
Department of Geophysics, Stanford University, MC 2215, Stanford, CA 94305 2USGS, 345 Middlefield Rd., Menlo Park, CA 94025 3Berkeley Seismological Lab, 215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760 *Currently at UC San Diego, La Jolla, CA 92092
1. Introduction
Stanford, in collaboration with UC Berkeley and the USGS, has maintained five high-precision ultra-low-frequency electromagnetic (ULFEM) recording stations along the San Andreas fault system since 2005. The sites are intended to monitor and record ULFEM signals in the Earth’s crust associated with seismic activity, should such signals exist. Each ULFEM site has three orthogonal coil magnetometers, and duplicate sets of orthogonal 100m horizontal electrode pairs, aligned with magnetic North and magnetic East. The MHDL site has only magnetic sensors. All sites collocated with a broadband seismometer. signal telemetry
other site equipment
Isolated power supply
Qua Dig
Hy
Magnetic field sensors Hz
nterra itizer cables to digital acquisition system
5-10 m buried cables
~100 m
er supply
Legend
electric field signal conditioner
Ex
A
electrodes
~100 m
38°N "
MHDL
Hx
magnetic sensors
pow
N
BRIB
37°30'N
A
"
Exy common
magnetic coils
A
JRSC
B
Ex
B
Electrodes
As our strategy for detecting possible EM anomalies related to earthquake processes relies on comparing measured EM signals across multiple sites, it is necessary to confirm that each magnetic sensor at each site responds similarly and accurately. A portable system placed on the ground surface above each buried magnetometer generates a time-varying magnetic field of known magnitude at several different frequencies in the bandwidth of interest.
The BSL Digitizer is an 8-channel, 24 bit device with a noise floor of 4 bits. The timing is given by GPS signal. The digitizer is linked by serial connection to a Linux-based single board computer to provide data storage, IP connectivity and operational software. The BSL Digitizer and Linux computer operate off 10 -14 VDC and consume < 10 W.
BSL Digitizer, 2min, 1Hz sampling
B
BSL Digitizer, 72hrs, 40Hz sampling
~100 m
figure not dra
wn to scale
In the summer of 2010, a newly designed data acquisition and logging system, called the BSL (Berkeley Seismological Lab) Digitizer, was prototyped and tested. The new system is intended to replace the ULFEM group’s currently used digitizers (Quanterra) and electric- and magnetic- field signal conditioners (EMI, Inc.) at a fraction of their cost, while maintaining signal fidelity and system noise at acceptable levels. Several issues related to the new hardware are apparent, most seriously a daily time drift related to a faulty internal timing mechanism. These issues are being addressed and are helping inform the development of the next-generation BSL Digitizer.
3. Coil Calibration
Presented here are results of the initial testing and benchmarking. The BSL Digitizer produces data very closely correlated with the existing Quanterra data, a positive sign for future implementation.
"
Batt.
Exy common
2. The BSL Digitizer
37°N
Ey
A
Faults
Ey
B
Schematic diagram of a typical site (from Karakelian et al., 2002)
"
EM stations
"
122°30'W
122°W 0
12.5
25
SAO
121°30'W
Map of the ULFEM sites (PKD not shown) (from Karakelian et al., 1988)
Strong evidence that ULFEM anomalies may be earthquake precursors was found during the 17 October 1989 Ms 7.1 Loma Prieta, California earthquake, when an unusual fluctuation in ultra-low frequency (0.01–10 Hz) magnetic signals was recorded (Fraser-Smith et al., 1990). Out ULFEM group hopes to verify whether or not similar signals occur prior to other earthquakes.
Quanterra Digitizer, 72hrs, 40Hz sampling
C
Spectral Plot, BSL Digitizer, 72hrs
Difference Plot
All data windows start at 00:00 on July 16, and show data from the N-S magnetometer channel. Note that in each time-domain plot, the DC shift is removed and the data are normalized to a range of +/-1. A: The data correlate well under visual examination, with spikes showing up in both graphs and longer-period trends matching closely. The peaks evident in the difference plot may well be due to the subsampling scheme used rather than an issue inherent to the digitizer.
Spectral Plot, Quanterra Digitizer, 72hrs
ULF anomalies recorded 10km from the Loma Prieta epicenter (half-hour signal averages on an east-west magnetic coil). Note the increase in activity two weeks prior to the main shock and much larger amplitudes three hours prior to the main shock (Fraser-Smith et al., 1990).
Hardware developments in 2010 will provide increased reliability and calibration for our stations. A new data collecting and logging system is under development, and its first field test is reported here. We also present a method for quickly and easily calibrating buried magnetometers.
References Fraser-Smith, A.C., Bernardi, A., McGill, P.R., Ladd, M.E., Helliwell, R.A., and Villard, Jr., O.G. Lowfrequency magnetic measurements near the epicenter of the Ms 7.1 Loma Prieta earthquake. 1990, Geophys. Res. Letts. v. 17, 1465-1468. Karakelian, D., Klemperer, S.L., Fraser-Smith, A.C., and Thompson, G.A., Ultra-low frequency electromagnetic measurements associated with the 1998 Mw 5.1 San Juan Bautista, California earthquake and implications for mechanisms of electromagnetic earthquake precursors. 2002, Tectonophys., v. 359. Merzer, M. and Klemperer, S.L. Modeling low-frequency magnetic-field precursors to the Loma Prieta earthquake with a precursory increase in fault-zone conductivity. 1997, PAGEOPH, v. 150, 217-248. Neumann, D.A., McPherson, S.-L., Kappler, K., Klemperer, S., Glen, J., and McPhee, D. Stanford–USGS Ultra-Low Frequency Electromagnetic Network: Status report and and data availability via the Web. EOS Trans. 2008, AGU, 88 (52), Fall Meet. Suppl., Abstract S53B-1824.
The circuit built to produce a sinusoidally varying current used to drive a wound air-coil is shown below. It is cheap, easy to make, and capable of shifting between several known frequencies. Actual sine-wave generation is controlled by the XR2206 IC chip, which is accurate to within 5%. Power Supply (Battery)
Circuit Board and Box Amp
Auto-Shifting Pot. Counter + Demux
Physical Coil
Signal Generation: Controlled by IC
Resistor Ladder
Block Diagram of Function Generator Circuit
Difference Plot
Spectral Difference Plot
B: The natural diurnal cycle of the EM field in the Earth’s crust is evident over this 3-day window. The obvious amplitude drift with periods >> 1hr could potentially be attributable to issues with either the BSL or the Quanterra digitizers, or, more probably, to differences between the magnetometers recorded by each digitizer. Although nominally identical, these coils have significantly worse noise performance and reduced output for periods > 100 sec data sets. In addition, the coils were subject to slightly different physical conditions at the field site, which may help explain the disparity. Major short-period spikes in the data still correspond between the two systems. C: The frequency spectrum of interest (100s to 20Hz). The data correspond well in the frequency domain, especially at lower frequencies. The same digital low-pass filter has been applied to both sets of data, but the Quanterra data is additionally filtered at a 20 Hz corner by an unknown scheme during sub-sampling.
4. Future Work BSL Digitizer:
- Hardware concerns related to the time delay and drift, as well as the problems with the on-board computer, will inform the next iteration of the BSL Digitizer prototype. By the end of November, a second field test should be under way. - A conversion scheme from machine counts to S.I. units for both digitizers should be established and verified. - In the second field test, each of the 8 data channels will be tested against each other to determine internal system noise. - At present we are assuming the bulk of observed differences are due to the BSL Digitizer being noisier than the Quanterra, which is 10x more expensive. However, the long-period drift (periods > 1hr) may be a function of the different magnetometers in use, as these were only intended to measure periods < 100s. - We need to decide what an acceptable noise level is for our continued ULFEM monitoring applications.
Coil Calibrations:
The calibration system is brought to a given field site and placed on the ground surface above each buried magnetometer. When activated, it is designed to automatically shift between seven different frequencies, transmitting each for about five minutes.
Signal Generation:
Timer
Quanterra Digitizer, 2min, 1Hz sampling
50 km
Data Collection:
Coil Construction:
The coil used to generate a magnetic field is hand-wound with 40 turns of 26-gauge magnetic wire. The magnetic field in teslas at the center of such a coil is: B
μ0 I N
Where I is current (amps), N is the number -7of turns, a is its × radius (meters), and μ̥ = 4 π 10 T*m/A
2a
For a coil placed a distance r off-axis, and distance x along axis from the midpoint of the coil: B x r, x B r r, x
μ0
I
2a
1
π
Q
μ0
I 2a
π
Q
E k E k
2
1
2
K k
Q 4 1
2
2
K k 2
where: r a; x a; x r; Q 1 ; k 4 Q; K(k) and E(k) are the complete elliptic integral functions of the first and second kind, respectively. Bx and Br are the along axis and radial components, respectively.
The geometry of this particular system is such that we are driving a magnetic field on the order of 100 nT at the coil center. This is well above the background noise level on the order of several to tens of nT.
- The next step for testing and calibrating our magnetometers is to pursue a direct winding project. By far the largest source of uncertainty in the measurements with our portable system is the simple geometry and depth to the buried magnetic sensor. This can be eliminated if a direct winding is installed around each sensor. - Each winding would be powered by a system similar to the one used here, and remotely triggered at regular intervals. - As at any magnetic observatory, we will disturb the measuring equipment as little as possible. Once we have a stable and reproducible method of wrapping a test coil around our magnetometers, we will eventually implement this across all our sites.
Results:
Our data from three sites, all using nominally identical sensor coils, shows that all three north-south magnetometers have response curves identical to within about +/-10%. The dominant source of these apparent differences is likely due to uncertainty in the precise location of our calibration coil with respect to the midpoint of the buried sensor (unkown to within a few decimeters). 350
N-S Components of JRSC, BRIB, MHDL JRSC MHDL BRIB
300 250 200 150 100 50
Q 4 2
Field data collection and testing at the BRIB and JRSC sites.
Magnetic Response (nT / Hz1/2)
1
0 -1 10
0
10 Frequency (Hz)
1
10
Response to the generated magnetic field in the N-S component magnetometers at three different locations.
Our portable system allows us to routinely demonstrate that our magnetometers are functioning as designed, and that any large changes in recorded signal level are external to our system.
Data Availability
All data are available via our website (Neumann et al, 2008): http://ulfem-data.stanford.edu
Acknowledgments
Major Instrument Funding: NSF- EAR-Earthscope 0346236 and NASA NNH08AH44I Undergraduate Research Funding: IRIS, NSF and Stanford VPUE Instrument Maintenance Assistance: Berkeley Seismological Lab Instrument Site Support: Jasper Ridge Biological Preserve
