Refraction Geophone Cluster Test - preliminary data and results
J. Louie, 10/24/97
http://www.seismo.unr.edu/ftp/pub/louie/reno/clts/clts.html
The data and analysis below show how lightweight single 8 Hz seismic refraction
geophones can be used to record frequencies below 5 Hz.
Such seismometers weigh just a few pounds each, and up to 48 can be recorded
simultaneously by the MSM Bison Galileo-21 instrument.
The success of these tests suggests that refraction geophone arrays can
record surface waves yielding S-wave velocity estimates to 30 m depths.
Equipped with 10 cm spikes, these vertical geophones require only approximate
hand leveling, shallow burial, and withstand rough handling.
Standard exploration seismometers, they contain a moving coil held by a leaf
spring that can travel a few
millimeters vertically relative to a magnet fixed to the spike, and thus
produce relative voltage amplitudes proportional to vertical ground velocities.
A crew of four can obtain a noise recording and pack up in about two hours.
Most of the effort is expended in handling the two 350 m long, 30 kg cables
that connect the seismographs to the recorder.
On July 16, 1997 J. Louie, R. Abbott, M. Herrick, and C. Mann performed
a cluster recording test on the UNR campus just north of the KNPB building,
and 30-40 m west of Virginia St.
The site was bladed and rolled soil and gravel fill.
24 geophones at a time were clustered side-by-side into a one meter square pit
30 cm deep, covered with soil, and tamped lightly.
13 records were recorded, seven of background noise, and four of 10 stacked
5 kg sledgehammer impacts to a steel plate, 10 m from the cluster.
Since the size of the geophone cluster is much less than the seismic wavelengths
at any frequency of interest here, the record from every geophone should be
exactly the same.
The cluster test shows the degree of amplitude, frequency, and phase response
variation between geophones.
All geophones and takeouts were numbered.
All records are 8 seconds long, with 2000 time-points at 0.004 s intervals.
The pre-amplifier analog low-cut filter could not be set at lower than 4 Hz.
The analog high-cut filter was set at 64 Hz.
The table below gives the parameters of each record that were varied.
Record Geophones Preamp Gain Source
1 1-24 20 dB Noise
2 1-24 20 dB Noise
3 1-24 20 dB Noise
4 1-24 20 dB Noise
5 1-24 20 dB Sledge impacts
6 1-24 20 dB Sledge impacts
7 25-48 20 dB Noise
8 25-48 20 dB Noise
9 25-48 20 dB Sledge impacts
10 25-48 20 dB Sledge impacts
11 25-48 0 dB Noise
12 25-48 40 dB Noise
13 25-48 60 dB Noise
Data Transfer and Conversion
The MSM Bison Galileo-21 recorder saves each record in SEG-2 format on its
internal DOS hard disk.
The SEG-2 files are copied to a Zip disk connected to the Bison (having
standard PC ports), which is then mounted on a networked PC for FTP to Sun
disks.
Our Bison is experiencing a firmware error that results in truncation of
ASCII header fields and their null separators in the header of each trace
in the SEG-2 file.
The error does not affect the instrument's own ability to display the records.
However, SEG-2 to SEG-Y data translation software from the Kansas Geological
Survey and
from the Colorado School of Mines cannot read records from our instrument.
This
C code
and compiled
Sun SPARC binary
have been modified from the Colorado School of Mines program to handle the
header field truncation.
The text file
``segykeyw.ord''
must be present in the folder with the SEG-2 files when the program is executed.
A
conversion script
runs the SEG-2 to SEG-Y conversion program for the 13 SEG-2 files to produce
a single file in
SEG-Y format.
This file has trace and some header data in 32-bit Sun binary float format,
instead of the IBM or IEEE float format specified for SEG-Y.
The conversion script also produces a file with 13 planes of 24 traces of
2000 time-points each (with dt=0.004 s), as an
unlabeled Sun binary float data volume.
Using the /rg/rms.c program, the RMS amplitude of each of the 13 records is
summarized in this
text file.
The /rg/tegain.c program was also used to trace-equalize the data by applying
a gain to each trace to produce a 0.02 RMS amplitude for between 2 and 6 s
record time, yielding another
unlabeled Sun binary float data volume for
plotting purposes.
A plot of the data with the trace-equalized amplitude keyed to gray level is
at left, and is also rendered in
PDF
and
PS
format.
Note that each of the 13 records appears as a horizontal band; record 1 is at the
top and record 13 at the bottom.
The geophone 25-48 records clearly contain a sign-reversed trace; this is
due to a cable connector reversal and does not affect relative frequency response.
Spectral Analysis
Given the small size and mass of the geophones, their shallow burial and
8 Hz principal frequency, and the 4 Hz lower limit on the pre-amp low-cut
filter, the spectral content of the data must be demonstrated. A
shell script
invokes the /rg/xfilter.c program to prepare summed power spectra of one or
more records.
The spectrum files are unlabeled binary float, 1024 elements long each.
The first element is the DC component; the last is the 125 Hz Nyquist component;
and df=0.12207 Hz/component.
The script computed the summed spectra of all 24 traces of noise records
4, and
12 and 13; and of impact records
5 and 6, and
9 and 10.
The /rg/log.c program takes the log value of each component, giving the
log summed spectra of noise records
4, and
12 and 13; and of impact records
5 and 6, and
9 and 10.
These log spectra files were concatenated into a single
binary file
for plotting with the /rg/traceplot.c program.
The results are plotted below, and are also available in
Adobe Illustrator 5.5,
PDF, and
PS formats.
The differences between the record 4 and records 12+13 summed spectra
are due to lack of FFT normalization in the /rg/xfilter.c program, to the
records 12+13 spectrum summing twice as many traces as the record 4 spectrum,
and to the much higher pre-amp gains used in recording records 12 and 13.
Noise appears between 5 and 60 Hz to fall as 1/f, while the hammer impacts are
adding energy between 25 and 75 Hz.
Some energy below 5 Hz is clearly above the power level of the spectra above
60 Hz, suggesting that very low frequencies of ground velocity are being recorded.
Low-Pass Filtering
Since low-frequency energy is being recorded, low-pass filtered records
were examined to test the coherency of the clustered geophones.
The /rg/xfilter.c program was used to derive another
unlabeled binary float file
of all 13 records, after Fourier-domain low-pass filtering employing a linear
filter response ramp from 25 to 30 Hz.
The variable-density plot at left is also available in
PDF and
PS formats.
Coherency between the clustered geophones below 25 Hz is easily visible
in wiggle-trace plots prepared with the /rg/traceplot.c program and
this parameter file.
At left below is the filtered noise record 4 (also as
binary float,
PDF, and
PS); at right is impact record 5 (also as
binary float,
PDF, and
PS).
Both show no phase distortion, even with their relatively low 20 dB preamp
gain setting.
The plots at left and below show the importance of using higher preamp gains
to record coherent energy below 5 Hz.
/rg/xfilter yielded a low-pass filtered version of the 13 records with a
linear filter response ramp from 5 to 6 Hz, as a
binary float file, plotted at left with a low
grayscale saturation amplitude (and in
PDF and
PS).
A plot with a higher saturation (or clip; not shown, available as
PDF and
PS)
demonstrates that higher amplitude recovery and coherence is possible below
5 Hz at the highest 60 dB preamp gain.
The wiggle-trace plots of 5 Hz low-pass filtered records show that much
although not all
coherency is lost at low frequencies if the preamp gain is not at least 40 dB.
Compare below the 20 dB noise record 4 (also as
binary float,
PDF, and
PS)
against the 40 db noise record 12 (also as
binary float,
PDF, and
PS).
Not shown but exhibiting the same effects are the 20 dB impact record 5
(available as
binary float,
PDF, and
PS),
and the 60 dB noise record 13 (available as
binary float,
PDF, and
PS).
Correlation Tests
Coefficients of correlation were computed to compare the first trace of
records 4, 5, 6, 9, 10, 12, and 13 against the 23 other traces in the
record. A
shell script for all records executes a
shell script for one record,
the text output of which then passes through a program
(C code and
Sun SPARC binary)
that finds the average correlation absolute value for each record.
Correlation results are given for each trace in text files for:
Raw data correlations,
25 Hz LP correlations, and
5 Hz LP correlations.
The following table gives representative correlation coefficients for
combinations of preamp gain and frequency range for cluster noise records.
The 20 dB correlations are taken from record 4, and are similar to the
correlations of all other 20 dB records, including impact records.
The 40 and 60 dB records are the noise records 12 and 13, respectively.
| Preamp Gain | 0-5 Hz | 0-25 Hz | Raw |
|---|
| 20 dB | 32.1 ± 10.2% | 94.4 ± 1.6% | 92.2 ± 1.6% |
| 40 dB | 92.9 ± 2.2% | 96.4 ± 3.4% | 96.3 ± 3.3% |
| 60 dB | 97.8 ± 2.0% | 97.2 ± 2.4% | 96.6 ± 2.4% |
The data above show that all instrument preamp gain settings except zero
can accurately record data above 5 Hz; including urban background noise and
nearby sledgehammer impacts.
A preamp gain of 40 or above will accurately record below 5 Hz, although raw or
higher-frequency data will show digitizer clipping from nearby impacts.
The sledgehammer must be placed closer than 10 m from the geophone to cause
physical clipping (or ``pin the seismometer to the stops''), as no clipping is
observed here at lower gain.
The records show poor but visible coherency below 5 Hz at 20 dB, suggesting that
any data recorded with similar instruments might yield low-frequency velocities.