GEM 2 Multi-Frequency EM Sensor
About the GEM-2
Terraplus is pleased to offer the GEM-2 - a portable, digital,
broadband electromagnetic sensor - for commercial sale. Combining
cutting-edge technology with simplicity of use, the GEM-2 is the finest
electromagnetic geophysical sensor for geological, environmental, and
geotechnical shallow-earth surveys. The GEM-2 package consists of the
sensor boom (ski), the console, a battery charger, and a shoulder strap.
It also includes a CDROM containing the manual and the operating
software - WinGEM - that runs on Windows.
The GEM-2 is most commonly used as a hand-held sensor. Whether the
operator is standing still, walking, or running, the instrument
accurately maps variations in electrical conductivity and/or magnetic
How Does the GEM-2 Work?
A new broadband electromagnetic sensor, GEM-2, is a hand-held, digital,
multi-frequency sensor. It operates in a frequency range of 330 Hz to 24
kHz, and can transmit an arbitrary waveform containing multiple
A frequently-asked question is the "Depth of Investigation." This is a
very complex question because the answer depends on many factors,
particularly on ground conductivity and ambient electromagnetic noise.
Based on many analyses and field data, we estimate the GEM-2 should be
able to see about 30-50m in resistive areas (>1000ohm-m) and about
20-30m in conductive areas (<100ohm-m). This figure assumes an ambient
noise level of 5ppm. The noise level is generally high in urban areas
and low in rural areas. For typical applications, we do not recommend
the GEM-2 for depths deeper than 50m.
The GEM-2 ski contains three coils: transmitter, bucking, and
receiver coils. For a frequency-domain operation, the GEM-2 prompts for
a set of desired transmitter frequencies. Built-in software converts
this into a digital â€œbit-stream,â€ which is used to construct the desired
transmitter waveform (Figs 1 and 2). This bit-stream represents the
instruction on how to generate a complex waveform that contains all
frequencies specified by the operator.
|Figure 1. A three-frequency transmitter waveform.
|| Figure 2. First 33 points of Figure 1.
The base period of the bit-stream for GEM-2 is set to 1/30th of a second
for areas having a 60-Hz power. The TX switches at 96 kHz and,
therefore, the bit-stream contains 3,200 steps within the period.
Through a Fourier transform of the transmitter current waveform above,
we obtain a power spectrum of the primary field, which shows each
GEM-2 Principle of Operation
A new broadband electromagnetic sensor, GEM-2, is a hand-held, digital,
multi-frequency sensor. The GEM-2 operates in a frequency range of about
300 Hz to 24 kHz, and can transmit an arbitrary waveform containing
multiple frequencies. The unit is capable of transmitting and receiving
any digitally-synthesized waveform by means of the pulse-width
modulation technique. Owing to the arbitrary nature of its broadcast
waveform and high-speed digitization, the sensor can operate either in a
frequency-domain mode or in a time-domain mode.
Depth of exploration for a given earth medium is determined by the
operating frequency. Therefore, measuring the earth response at multiple
frequencies is equivalent to measuring the earth response from multiple
depths. Hence, such data can be used to image a 3-D distribution of
subsurface objects. Results from several environmental sites indicate
that the multi-frequency data from GEM-2 is far superior in
characterizing buried, metallic and non-metallic targets to data from
conventional single-frequency sensors.
Of many geophysical exploration techniques, the electromagnetic (EM)
method provides significant advantages for shallow geophysical
exploration. The first sensor, Model AEM-1, a helicopter-towed unit
weighing about 250 pounds, was built for the U.S. Navy and successfully
flown in 1988 for airborne bathymetric surveys over the shallow ocean.
The first hand-held version, GEM-1 completed in 1992, weighed about 12
pounds and was used for many environmental site characterization surveys
to detect landfills, unexploded ordnance, and buried drums containing
GEM-2, the most recent portable version, completed in early 1995, is
shown in Figure 1. The sensor, weighing about 9 pounds, operates in a
frequency band between 90 Hz to about 24 kHz. Its built-in operating
software allows a surveyor to cover about one acre per hour at line
spacing of five feet. Along a survey line, the data rate is about two
per foot, resulting in about 20,000 data points per acre per hour. Such
portability, survey speed, and high data density are important
requirements for geophysical surveys at environmental sites.
|Figure 1. GEM-2, a multifrequency EM sensor, in operation.
Advantages of a broadband, multifrequency, EM sensor are obvious. The idea
of using multiple frequencies stems from the so-called â€œskin-depth,â€
also known as the depth of exploration, which is inversely proportional
to frequency: a low-frequency signal travels far through a conductive
earth and, thus, "sees" deep structures, while a high-frequency signal
can travel only a short distance and thus, "sees" only shallow
structures. Therefore, scanning through a frequency window is equivalent
to depth sounding. Figure 2 shows a nomogram from which one may
determine the penetration depth for a given frequency (Won, 1980).
Skin Depth Nomogram To determine depth of penetration, connect the
soil/rock type on the left to the frequency of operation on the
right. The intersection in the middle indicates the depth of
penetration. From Won (Geophysics, 1980).
Depth sounding by changing the transmitter frequency is called â€œfrequency
sounding,â€ which measures the target response at many frequencies in
order to image the subsurface structure. Because the method involves a
fixed transmitter-receiver geometry, the sensor can be built into a
single piece of hardware, such as GEM-2; as a result, it produces
extremely precise, sensitive, and thermally stable measurements. In
contrast, depth sounding by changing the separation between the
transmitter and receiver is called â€œgeometrical sounding,â€ which usually
requires multiple operators tending separate coils connected by wires
and measuring consoles. Maintaining a precise coil separation is
difficult and, therefore, some measurements (e.g., in-phase components)
are often abandoned. For shallow surveys, the frequency sounding method
offers high spatial resolution, survey speed, light logistics, and data
2. GEM-2 Operating Principles
Figure 3 shows the electronic block diagram of GEM-2. The sensor contains
a transmitter coil and a receiver coil separated by about 5.5 feet. Such
geometry is called â€œbistaticâ€ configuration. It also contains a third
â€œbucking coilâ€ that removes (or bucks) the primary field from the
receiver coil. All coils are molded into a single board (dubbed â€œskiâ€)
in a fixed geometry, rendering a light and portable package. Attached to
the ski is a removable signal-processing console.
|Figure 3. GEM-2 electronic block diagram.
A PC is to be connected to the GEM-2 console through an RS-232 cable for
uploading the operational parameters or downloading the data using
WinGEM, a Windows-based program. For a frequency-domain operation, the
program prompts for a set of desired transmitter frequencies. Built-in
software converts these frequencies into a digital â€œbit-stream,â€™ which
is used to construct the desired transmitter waveform for a particular
survey. This bit-stream represents an instruction on how to control a
set of digital switches (called H-bridge) connected across the
transmitter coil, and generates a complex waveform that contains all
frequencies specified by the operator. This method of constructing an
arbitrary waveform from a digital bit-stream is known as the pulse-width
The base period of the bit-stream for GEM-2 is set to 1/30th of
a second for areas having a 60-Hz power supply as does the U.S. The
period is 1/25th of a second for 50-Hz areas, as in Europe and Japan.
For an H-bridge switching rate at 96,000Hz, for instance, the bit-stream
contains 3,200 steps within the base period. Any integral number of the
base period may be used for a consecutive transmission in order to
enhance the signal-to-noise ratio.
In Figure 4a, we show an example transmitter current waveform, generated
by a bit-stream designed to transmit three frequencies, 90Hz, 4,050Hz,
and 23,970Hz. The software that generates the bitstream for a given set
of operator-specified frequencies takes into account the built-in
electronic parameters, such as the resistance and inductance of the
transmitter coil. The graph shows the current waveform for the bit
sequence from 1 to 1,067, the first third of the base period, 1/30th of
a second. Figure 4b depicts the waveform details for the first 33 bits,
showing the current flow in the transmitter coil. Each bit in this case
lasts for 1/96,000 seconds. Figure 4c shows the amplitude spectrum of
the transmitter current waveform of Figure 4a. Note that the transmitter
current decreases logarithmically with frequency. The maximum current
(peak to peak) for the present transmitter is close to 10 amperes,
corresponding to a dipole moment of about 3 A-m2.
Figure 4a. An example transmitter current waveform (shown
only one-third of 3,200 points) generated by a 3-frequency bit
Figure 4b. Enlarged section of Figure 4a between the bit
sequence 1 to 33.
Figure 4c. Amplitude spectrum of the transmitter current
waveform shown in Figure 4a.
GEM-2 has two recording channels: one from the bucking coil (called the
reference channel) and the other from the bucked receiver coil (called
the signal channel). Both channels are digitized at a rate of 48,000 Hz
(half the transmitter rate) and at an 16-bit resolution. This produces a
1,600-long time-series per channel during a base period. In order to
extract the inphase and quadrature components, we then convolve (i.e.,
multiply and add) the time-series with a set of sine series (for inphase)
and cosine series (for quadrature) for each transmitted frequency. This
convolution renders an extremely narrow-band, match-filter-type, signal
detection technique. A 3-frequency operation, shown in Figure 4a, with a
transmitter duration of 33 milliseconds, typically requires a total of
about 60 milliseconds to complete the entire measurement cycle per data
location. Increasing the number of frequencies would reduce the
measurement cycle. A single computer in a DSP chip coordinates all
controls and computations for both transmitter and receiver circuits.
GEM-2 may also be used to measure the background environmental noise
spectrum up to 24kHz. This is obtained from the signal-channel
time-series at a typical location within a specified survey area, then
computing its entire Fourier spectrum at an interval of the base
frequency (30 Hz). Using the environmental noise spectrum, the operator
can safely avoid locally-noisy frequency bands.
3. Basic Measurement Unit
The in-phase and quadrature data derived through the convolution are
converted into a part-per-million, or ppm, unit defined as:
These ppm values are the raw data logged by GEM-2. It is obvious
that the ppm unit defined by Eq. (1) is sensor-specific and has little
physical meaning. GEM-2 allows up to about 50,000 data points before
down-loading to a PC through an RS-232 protocol.
All parameters required for the ppm computation, such as the sensor
output in free-space (simulated by hanging GEM-2 from the top of a tall
tree), amplifier characteristics of the two receiving channels, and the
coil geometry, are stored in GEM-2 for real-time use.
In most shallow geophysical surveys, the ppm data generated by
GEM-2, often plotted into a contour map for each frequency, are
sufficient to locate buried objects without going through elaborate
processing or interpretation. One can also estimate the target depth
from the data obtained at multiple frequencies.
This modus operandi is the most common â€œbump finderâ€ survey,
which is appropriate and productive where there are numerous shallow,
small, nondescript targets and the survey objective is to find as many
targets as possible. The goal is not to determine a detailed geometry
for each object, given typical time constraints and the large quantities
of objects to be detected. In such a survey, the in-phase and quadrature
ppm data are sufficient to indicate the location, size, and depth of a
â€œbumpâ€ without converting the data into any other more physically
Figure 5 is shown as an example; we look for a buried pipe and our main
interest in this case is its location. This figure shows the GEM-2
in-phase response at 7,290 Hz over a known stainless-steel pipe of
18-inch diameter, buried at a depth of approximately 30 feet. A magnetic
survey failed to detect the pipe, presumably because it is made of
stainless steel, a non-ferrous metal. In this example, the plot showing
the ppm response is sufficient to locate the pipe. The survey over this
pipe included seven frequencies, and the response was highly dependent
on frequency. For example, the pipe was not recognizable at around 2 kHz
or 12 kHz.
|Figure 5. GEM-2 7,290 Hz in-phase data over an 18-inch
diameter stainless steel pipe buried at 30 feet below surface.
4. Conversion to Apparent Conductivity
Since the in-phase and quadrature ppm data contain all information on the
measurement geometry, they can be the raw input for any inversion
software. Traditionally, however, EM data are displayed in â€œapparent
conductivityâ€ by imagining that the earth below the sensor is
represented by a homogeneous and isotropic half-space. While the earth
is heterogeneous with regard to geologic variations, it can be
represented by an equivalent homogeneous half-space that would result in
the same observed data.
GEM-2 measures the secondary field from the earth (and buried
objects therein) at frequencies specified by the operator. When the
field is normalized against the primary field at the receiver coil, it
is called the mutual coupling ratio (Q), which, for horizontal coplanar
mode (or vertical dipole mode), can be written as:
The sensor geometry with respect to the earth is shown in Figure
6. The kernel function R corresponding to a uniform half-space is:
Hs: secondary field at receiver coil,
Hr: primary field at receiver coil,
r: coil separation,
h: sensor height,
J0: zeroth-order Bessel function,
f: transmitter frequency (Hz),
Âµ: magnetic susceptibility, and
s: earth conductivity.
We note that the ppm unit defined by Eq. (1) is the same as Q multiplied
by one million. GEM-2 can be configured to a vertical coplanar mode (or
horizontal dipole mode) by simply turning it 90 degrees about the ski
axis (Figure 7).
The mutual coupling ratio for this is: (4)
where J1 is the first-order Bessel function. The integrals in s. (2) and
(4), known as the Hankel transform integrals, can be computed by the
linear digital filter method with known filter coefficients for a fast
digital convolution (Kozulin, 1963; Frischknecht, 1967). We notice from
the above equations that conductivity and frequency appear as a single
product. For multifrequency data, therefore, Eqs. (2) and (4) provide
the relationship between the ppm unit and the conductivity-frequency
|Figure 6. GEM-2 in horizontal coplanar coils
||Figure 7. GEM-2 in vertical coplanar coils configurations.
Figures 8 and 9 illustrate the computed in-phase and quadrature responses
over half-space for the GEM-2 horizontal and vertical coplanar coils. We
assume for this example the sensor height at 1 meter, typically waist
level for a surveyor. In essence, the observed ppm value (y-axis)
determines the conductivity-frequency product (x-axis), which is then
divided by the transmitter frequency to obtain the half-space
||Figures 8 & 9. GEM-2 interpretation charts (in-phase and
quadrature) for converting ppm data to earth conductivity.
As an example, Figure 10a through 10d show the GEM-2 apparent conductivity
data, in units of millisiemen/m (mS/m), over a 6-acre trench complex in
the southeastern U.S. Materials, including radioactive waste, were
"systematically" buried in many parallel trenches. For comparison,
Figure 10e shows, for the same area, the total-field, magnetic anomaly
map, while Figure 10f shows the total-field, vertical magnetic gradient.
For the magnetic data, we employed a cesium-vapor magnetometer
(Geometrics G-858G) with two sensing heads vertically separated by 30
inches. Magnetic anomalies are inherently dipolar in nature and, thus, a
target is commonly located at the slope, rather than the peak, of an
anomaly. This problem of target location renders the magnetic anomaly
hard to interpret, particularly for a target made of many individual
items such as drums and cans buried in these trenches. In contrast, it
can be shown theoretically by a forward modeling that a GEM-2 anomaly is
almost monopolar centered directly above the target and, consequently,
is easier to interpret than dipolar magnetic anomalies.
||Figure 10a through 10d
Figure 10g is from an old facility engineering drawing that supposedly
shows the trench locations. We do not know whether this old drawing was
intended to show the design plan before, or the â€œas-builtâ€ map after,
the trench construction. Regardless, it is obvious that the trenches
implied by the GEM-2 data are significantly dissimilar from those
indicated by the old map. In the end, we concluded that the GEM-2 data
best depicts the current trench distribution; therefore, the geometrical
boundaries determined from the GEM-2 data were used to compute the waste
volume within the trenches. Each map, depicting the apparent
conductivity derived from the in-phase and quadrature data at 1,350 Hz
and 7,290 Hz, shows a slightly different picture of the trenches.
Presumably, the low-frequency data indicates the deep trench structures,
while the high-frequency data indicates the shallow trench structures.
For plotting convenience, each conductivity map, Figures 10a through
10d, has an average conductivity value (noted on the top of the color
scale bar) removed in order to balance the color distribution. The
conductivity values may contain a constant offset resulting from
imprecise calibration of the GEM-2 free-space response; the problem is
Of the many geophysical sensor technologies, the EM method provides
significant advantages for shallow environmental characterization.
Unlike seismic or ground-penetrating radar methods that involve heavy
logistics and labor-intensive field work, GEM-2 requires only a single
operator, does not touch the earth (thus, is less intrusive), and can
operate at stand-off distance. An instrument like GEM-2 is ideal for
many environmental and geotechnical applications including mapping
underground storage tanks, landfill and trench boundaries, certain
contaminant plumes, and buried ordnance. In addition, GEM-2 has
applications for finding shallow orebodies for the mineral exploration
Despite many compelling advantages, the broadband EM method has not
received sufficient attention from the geophysical community. With the
advent of digital, multifrequency data, we have opened a new dimension
in data quality and quantity for imaging and characterizing buried
subsurface features. GEM-2 is only the beginning of a new generation of
many broadband EM sensors.
GEM-2 Technical Specifications
Bandwidth: 330 Hz to 24 kHz
Frequency Domain: Single Frequency, Multiple Frequencies,
Maximum Sampling Rate: 30 Hz
Removable Ski: Sealed Kevlar or fiberglass foam, 2kg
Console (CPU, display): 10 x 25 x 5cm anodized al box, 2.5 kg
Internal Rechargeable Battery: 13.2 VDC
Coil Configuration: Horizontal or Vertical Coplanar
Output: Inphase and Quadrature in ppm at all specified
Maximum TX moment: 3 Ampere square meters
Program upload and data download from a PC: WinGEM
PPM to conductivity conversion program
Data processing and display software
Real-time data output option for GPS integration
Layered earth inversion program
Frischknecht, F.C., 1967, Field about an oscillating magnetic dipole
over a two-layer earth and application to ground and airborne
electromagnetic surveys, Quart. Colorado School of Mines, v. 62, no.
1, pp. 1-370.
Kozulin, Y.N., 1963, A reflection method for computing the
electromagnetic field above horizontal lamellar structures,
Izvestiya, Academy of Sciences, USSR, Geophysics Series (English
Edition), no. 2, pp. 267-273.
I.J. Won, 1980, A wideband electromagnetic exploration method - Some
theoretical and experimental results, Geophysics, Vol. 45, pp.
I.J. Won, 1983, A sweep-frequency electromagnetic exploration method,
Chapter 2, in Development of Geophysical Exploration Methods-4,
Editor; A. A. Fitch, Elsevier Applied Science Publishers, Ltd.,
London, pp. 39-64.