A New Generation of Superconducting Gravimeters

Bernd Richter ¹⁾ and R. J. Warburton ²⁾

¹⁾ Institute for Applied Geodesy, Frankfurt / Main, Germany
²⁾ GWR Instruments, Inc., 6264 Ferris Square, Suite D, San Diego, CA, USA

Published in: Proceedings of the Thirteenth International Symposium on Earth Tides

Edited by B. Ducarme & P. Paquet,

Brussels, 1998

1. Introduction

In 1980 the first superconducting gravimeters (SGs) became commercially available and since then they have been operated very successfully worldwide. These instruments provide long term observations series which are essential to improved understanding of the time-dependent gravity field, especially causes of long-term gravity variations and the influence of environmental effects on gravity (e.g. Crossley et al. 1995, Ducarme et al. 1997, Elstner et al. 1993, Goodkind 1990, Peter et al. 1995, Richter 1984). The excellent results obtained result from the very low noise and drift characteristics of the SG. However, even with the high quality of the time series there are still gaps, offsets and noise in the registration due to strong environmental effects like earthquakes; man-made disturbances from instrument maintenance and liquid helium transfer; and limitations in the long term stability of the electronic components.

Whereas small gaps can be closed due to the low and constant drift behavior, offsets are not well known in magnitude and duration especially when they occur during periods of high geophysical noise. Experience with analysis of data has shown that apparent gravity (or drift) variations depend upon assumptions used to model and remove offsets (Harnisch et al. 1997). Several long term experiments were performed at Bad Homburg, Germany (1995-1996), Miami, Florida, USA (1989-1990) and Boulder, Colorado, USA (1993-1994) where two superconducting gravimeters were operated side-by-side for at least one year to study instrumental effects (Klopping et al. 1995, Richter 1987, van Dam et al. 1998). These experiments show that instrumentally induced disturbances are usually seen only in one instrument so that the undisturbed signal of the other instrument can be used to detect and correct any offsets or drift anomalies that occur. In spite of the success of these experiments, it is obviously too expensive and time consuming to operate two instruments at every location. Nonetheless, these observations spawned the idea to develop a new "dual sphere" superconducting gravimeter (DSG) which would combine the advantage of two in-situ registrations and the possibility of more simple and cost effective operation.

II. Description of the dual sphere superconducting gravity meter

The dual sphere superconducting gravity meter (DSG) is similar in construction to the single sphere superconducting gravity meter (SG) with the addition of a second sensor placed approximately 20 cm above the first. As shown in Figure 1, each gravity sensor contains a hollow niobium sphere with its vertical displacement measured by a capacitance bridge. Unlike earlier sensors which used only 2 coils, this prototype DSG used 3 superconducting magnet coils to levitate each sphere. The three coil design allows the operator to adjust both the magnetic gradient and the vertical null position of the magnetic gradient so that it is in precise agreement with the null of the capacitance displacement sensor. This allows the two sensors to be magnetically matched more precisely.

The magnet coils, feedback coils, and thermal switches are wound and mounted on a single cylinder of copper. The lower sphere capacitance bridge (Unit 1) is inserted from the bottom of the cylinder and anchored to an internal mounting flange. The upper bridge (Unit 2) is inserted from the top. The volumes containing the two spheres are evacuated, back-filled with helium gas and sealed with a copper flange and indium O-ring. Three sections of an Al cylinder attach to the midpoint of the Cu cylinder and mount it rigidly to an intermediate plate placed below the vacuum can lid. The intermediate plate is attached to the vacuum can lid using three G-10 posts which thermally isolate it from the vacuum can lid and helium bath. The Cu block is temperature regulated using a Ge thermometer thermally anchored at the midpoint of the block and a resistive heater placed nearby. In this configuration, the conductive heat flow between the sensor and the helium bath goes through the temperature controlled midpoint of the sensor. This minimizes thermal gradients in the sensor.

Two items complete the basic DSG. A niobium shield surrounds the sensor. Its open upper end slides past the sensor, the sensor mount, and the intermediate mounting plate. The closed lower end of this shield is centered by and attached to a post mounted concentric to the lower end of the DSG sensor. The circumference of the shield is slightly constricted where it mates with the intermediate plate. Therefore, it must be pressed into place which insures that a good thermal contract is made between the shield and body. A vacuum can encloses the shield and sensor. It is sealed to the vacuum can lid with an indium O-ring and evacuated. At the helium bath temperature (4.2 ^OK), helium gas is removed using charcoal getter while all other gases freeze out. Therefore, a hard vacuum is attained at cryogenic temperatures which enables extremely good temperature regulation.

III. Operation of the dual sphere superconducting gravity meter (DSG)

The DSG was first tested in a 60 liter laboratory dewar. The dewar was supported at three equilateral base points and leveled by adjusting two micrometer heads with respect to the third fixed point. Two active thermal levelers placed below the micrometers held the DSG at its tilt insensitive position during operation. The dewar/thermal levelers were raised off the ground using granite blocks. This additional clearance was used to apply gravity and gravity gradient signals to the DSG by inserting and removing a mass beneath the dewar. The mass was supported on a hollow Al cylinder and placed on a small cart guided by rails. Periodic signals were applied to the DSG by moving the cart in and out beneath the dewar.

The DSG operates with two sets of gravity electronics with one dedicated to Unit 1 and one to Unit 2 so that each of the two sensors can be operated as a single sphere SG. The outputs of these electronics are filtered and then sampled by a data acquisition system in the standard method used with the single sphere SG. After storing the data, the difference signal can be generated after correcting for the gain difference between the two units. The difference signal in m Gal can be converted to a gravity gradient signal in Eötvös by multiplying by a factor of 50, since the spacing between the sensors is 0.2 m and 1 Eötvös = 0.1 m Gal/m.

A. Tilt alignment and sensitivity

When an ideal gravimeter is tilted by an angle 2 from the vertical, it measures the component of gravity, gcosq , along its axis. Therefore, to lowest order, the ideal response is - ½ g q ² and the gravimeter will read a maximum value when aligned with the vertical. For the superconducting gravimeter the tilt response is quadratic for small tilt angles but, due to the magnetic field geometry supporting the sphere, it is of opposite sign. The tilt sensitivity of the instrument is a very important parameter because it measures the extent to which instrumental tilts will introduce spurious signals. To reduce this effect, the instrument should be adjusted near the minimum of the tilt function to decease its tilt sensitivity D g/D q .

The masses of the test spheres, magnetic shielding, coil winding and machining differences all produce magnetic asymmetries that differ between the two sensors in the DSG. These affect both the coefficient and null point of the individual sensors, so that the individual sensors have different tilt minima q ₁ and q ₂. The sensors individual dependencies on tilt can be expressed as: g₁ = C₁(q -q ₁)² and g₂ = C₂(q -q ₂)². In the initial tests runs the difference in the null positions, q ₁ and q ₂, of Unit 1 and Unit 2 was about 3.4 mradians. Also, tilting excited orbital modes with periods of 100 and 52 seconds on Units 1 and 2 respectively. The presence of the modes, and their slow decay, made the tilt nulling process slow and tedious.

To solve the alignment and mode problems, two sets of concentric side coils were added to the instrument. One set is wound with normal (Cu) wire and consists of 4 coils surrounding the sphere such that they produce a quadrapole field. This field is applied to the sphere when it is cooled from its normal to superconducting state, so that the field remains trapped in the sphere after the side coil currents are turned off. The motion of the trapped field in the nearby normal metals usually damps the orbital modes in a few seconds. In this instrument, Unit 1 was completely damped, however, Unit 2 had several modes remaining between 6 to 11 second periods. We discuss later how the presence of these modes introduced noise and limited the precision of the DSG.

The second set of coils, which are used only on Unit 2, are superconducting coils placed orthogonal to each other. They are shorted with superconducting/normal heat switches, and have current leads to room temperature. These coils are activated only after the spheres have been levitated and the magnetic force gradients adjusted. Currents trapped in these coils apply small horizontal forces to the sphere and are used to align the tilt null of Unit 1 with that of Unit 2. The tilt sensitivity to the coil is 65 :radians/mamp and the tilt nulls were aligned to within 10 :radians using this technique.

Figure 2 shows a plot of dg/dq versus q for the two individual sensors and the difference tilt signal. It can be seen that the tilt coefficient, C₁ = 8.3 x 10^-4 :gal/:rad², is about 30% smaller than C₂ = 10.8 x 10^-4 :gal/:rad². This leaves a difference tilt coefficient C_D = 2.4 x 10^-4 :gal/:rad². Therefore, the tilt sensitivity of the gradient signal is reduced by a factor of 3 to 4 times from the gravity meters.

Ideally, we want to equalize C₁ and C₂ so that the tilt sensitivity of the gradient is very small. For example, if C₁ and C₂ were within 1%, the DSG gradient tilt sensitivity would only be 1% of the gravity sensitivity. As stated above, the tilt coefficient is determined by weak horizontal restoring forces and on the mass of the sphere, the placement of the sphere with respect to the upper two coils, and the magnetic shielding. In future work, we will continue to equalize these parameters to further decrease the tilt sensitivity of the gradient. At this time, however, the noise from tilt is not the factor currently limiting instrumental noise. The DSG has two tiltmeters mounted above the vacuum can. Tilt signals feedback to the thermal holds the tilt within 0.2 :rad of its set point. With the set point adjusted to within 10 :rad of the DSG tilt null, the tilt noise produced D g_TILT = C_D x 0.2 x 10 :rad² = 4.8 x 10^-4 :gal. Expressed as a gravity gradient, this corresponds to D g_TILT = 0.024 Eötvös.

B. Gravity gradient measurements

1. Response to a nearby mass

To measure the signal to noise ratio of the superconducting gravity gradiometer , we moved a small mass beneath the dewar on a cart. Figure 3 shows the response of the DSG to the mass being rolled under the dewar and removed at approximately 10 minute intervals. To produce this record, each gravity meter was sampled individually at 20 second intervals after analog filtering through a lowpass filter with a corner frequency, f_C = .02 Hz (T_C = 50 sec). This is a six pole analog filter with a high frequency attenuation rate 120 dB per decade. It was designed specifically to attenuate microseismic signals with periods of roughly 6 and 12 seconds. After recording the gravity meter outputs in volts, each was converted into m Gal using a calibration previously determined by fitting longer records to earth tide signals. Differencing these records produces the output data shown. The step function produced by the mass is clearly observed on the data. The mean value of the measurement is 0.22 m Gal (11.1 Eötvös) with a standard deviation F = 0.01 m Gal (0.55 Eötvös). This is in good agreement with the gradient calculated based on the size of the mass and its distance from the center point between the sensors.

2. Oscillator and transformer stability

At low frequencies with periods of a few hours to days, we observed noise on the DSG. We proved that most of this was caused by temperature dependencies in the individual drive references and transformers. Figure 4 shows the effect of temperature over a 17 hour long record. In this record, a heat lamp was used to preferentially heat only one of the two sets of control electronics during the first two hours of the record. After this, the electronics cooled with the room overnight. An extremely good correlation between temperature and the difference signal between the two sensors was observed with a coefficient is about 0.3 m Gal/^OC .

We hypothesized that temperature induced variations in the output of the drive transformer secondaries effected the ratio of the drive on the upper and lower capacitor plates. This directly changes the output of the measured signal. Also, since the reference voltage produces an electrostatic force on the sphere, instabilities in the reference voltage will cause spurious changes in gravity. To confirm these observations, we operated both sensors with a common ultra-stable reference and a drive transformer with an extremely small temperature coefficient. Operating in this mode, no temperature induced signal was observed on the gradient for a 10 ^OC applied temperature variation. With this change, the sensor stability was about +/ 0.04 m Gal during a 17 hour test.

After completing the experiments described above, several changes were made to improve the gravity card electronics. These included: a new high precision reference oscillator; and improved low temperature dependent drive transformer, a high current transformer driver; and an improved 4 layer printed circuit board with shielded input stage and shielded transformer. Figure 5 shows typical noise and temperature performance for the new GWR gravity boards. Two traces show the raw 1 sec data from Unit 1 and Unit 2. In this test, the temperature of the gravity card used with Unit 1 was raised about 10 ^OC as seen in the temperature trace. The difference signal (U1-U2) shows no response from the temperature changes. Based on these data, the temperature coefficient is less than 0.1 mV/^OC » 10 nGal/^OC. This is an improvement of more than 30 when compared with the results shown in Figure 4.

IV. Removal of Offsets

As discussed earlier, the performance of the single sensor SG can be degraded by the occurrence of small spontaneous offsets (or "tares") in its signal. These offsets can be partially removed by post processing the data, however, the precision is limited by the presence of geophysical noise from seismic and atmospheric sources. Therefore, residual offsets interfere with interpretation of long term records and they also add noise at higher frequencies of geophysical interest.

Comparison of the two signal from the DSG allows the user to easily identify any offsets that occur and to remove them with a precision of better than 0.05 m Gal. Precise identification of offsets (both time location and magnitude) is made possible by the fact that most geophysical signals are common mode and are removed in the difference signal. A typical analysis of an offset and its removal is shown in Figure 6. The two traces U1 and U2 show the raw signals (1 sec sampling) from the two sensors and U1-U2 shows the raw difference signal and the presence of a 1.9 m Gal offset. The trace U1-U2-Offset shows the raw data after removal of the offset. Finally the Filtered trace shows the I1-U1-Offset signal after filtering with a 1 hour time constant. The standard deviation for U1-U2-Offset is » 0.1 m Gal and the standard deviation on the Filtered trace is » 0.03 m Gal. As seen in Figure 8, precise measurement of the offset size is limited only by electronic noise or true gradient signals. Since the offsets are due to random flux jumping in the superconducting sphere, coils or shield, it is extremely unlikely that two will occur in both sensors at the same time. Presently, three DSGs have been tested and operated and no coincidental offsets have yet been observed.

As described above, the difference signal makes it easy to identify any offsets greater than 0.05 m Gal. However, identification of the unit where the offset occurred can be more difficult, especially for small offsets. Offsets greater than 1 m Gal are usually easy to observe on a single unit by looking at a tidal residual signal or a bandpassed filtered signal. However, even offsets this size can be masked by the simultaneous occurrence of an earthquake or extreme atmospheric variations. Future work will focus on developing algorithms and data processing techniques to positively identify the unit responsible for small offsets so that they can be removed from the correct data set.

V. Ultra Long Holdtime Dewar refrigeration system

The ultra long holdtime dewar (ULHD) refrigeration system uses a dewar that is similar to the compact dewar. Its neck, however, is redesigned to interface with a cryocooler capable of obtaining temperatures below the vaporization point of liquid helium. The complete system is shown in Figure 7. It uses a KelKool 4.2 GM cryocooler manufactured by Leybold Vacuum Products Inc. This Gifford-McMahon type cryocooler uses a mechanically driven piston and offers superior cooling performance, as well as greatly reduced noise and vibration compared to previous gas driven cryocoolers. This two stage cryocooler delivers a maximum cooling power of 50 Watts at 50 ^OK to its upper stage and 0.5 Watts at 4.2 ^OK to its lower stage. The upper and lower stages of the cryocooler are aligned with the outer and inner shields of the dewar and GWR uses a proprietary design to gas couple these two cryocooler stages to the shields. This makes it extremely simple to remove the coldhead from the dewar when servicing or exchange is required. Normally the lower stage is operated below the vaporization temperature of helium so that the boiled off helium gas is re-condensed and drips back into the storage belly. Therefore, during normal operation the system consumes no liquid helium and will operate indefinitely.

The cooling capacity of the refrigerator is determined by many factors but most strongly by the rate at which the displacer cycles inside the coldhead. In the Leybold design, the displacer is driven mechanically by a stepper motor and yoke; and its speed can be controlled by adjusting the speed of the stepper motor. With the addition of a pressure sensor and a micro-controller the pressure of the dewar can be set and held at a constant value by varying the cooling power within a feedback loop. Figure 8 shows a schematic of this feedback loop The pressure is sensed by an Ashcroft pressure transducer and fed into the Tern micro-controller. In response to a high pressure (with respect to the set point) the controller increases its output frequency to the stepper motor controller which increases the coldhead speed and hence the cooling power. If the pressure is too low, the controller decreases its output frequency and decreases the cooling power. The system parameters of the feedback loop are programmed into the micro-controller.

Figure 9 shows an example how the controller responds after the system has been disturbed. At the start of the record, the dewar pressure is being controlled around the set point slightly above 93 Kpa. At about 3.5 hours the system is disturbed by venting the dewar pressure. In response the speed of the coldhead is increased to its maximum value of 140 rpm (cycle or revolutions per minute) which causes the dewar pressure to drop rapidly. After some delay, the coldhead speed also drops rapidly until the pressure enters the "proportional" band. If the pressure had decreased to less than 92 Kpa, the coldhead speed would have been decreased to about 10 rpm. Near the equilibrium point, which is re-established at about 13 hours, the coldhead speed is restricted to operate between 22 and 88 rpm. The required feedback characteristics are complicated by the long and asymmetrical response of the helium bath to pressure and by cooling power variations in in the coldhead at constant rpm.

The Leybold coldhead is designed to operate at its maximum capacity at 140 rpm. At this operating speed, the manufacturer specifies that the coldhead has a life expectancy of more than 10,000 hours (1.1 years). During normal operation in the ULH dewar, the coldhead operates with a mean cycle time of 30 rpm. It is expected that this will increase the life expectancy to more than 4 years. The compressor maintenance schedule requires exchange of the oil adsorber module at 2 year intervals. GWR supplies the compressor with a water refrigeration system which dissipates heat generated by the compressor to a remote location. This water refrigerator is a compressed freon system that cools the water to acceptable limits as required by the high performance system.

VI Conclusions

A "dual sphere" superconducting gravimeter is being developed which contains two independent gravity sensors in the same cryogenic container. By simply differencing the signals from these sensors, one can detect any offsets produced by magnetic flux "jumps" that are larger than 0.05 m Gal. Offset detection is no longer limited by the ambiguity of removing "real" gravity signals nor by sources of geophysical noise that may overlap and hide an offset. For offsets greater than 1 m Gal it is straightforward to assign the sensor that produced the offset. For smaller sub-m Gal offsets, however, data processing techniques need to be developed to rapidly identify and assign the source sensor.

Concurrent with this work the superconducting gravimeter’s control electronics are being improved. The focus is to decrease electronic noise, eliminate sources of electronic offsets, improve long-term stability and to simplify user operation. As reported herein, the temperature sensitivity of the gravity control card is decreased by more than a factor of thirty. This work is now being extended to optimize the gravity preamplifier and capacitance bridge network, and the temperature and tilt control electronics.

A new ultra long holdtime dewar is being developed using a 4 ^OK refrigeration system. The 4 ^OK cryocooler recondenses the helium gas as it boils off from the liquid helium bath stored inside the dewar. In this manner, the dewar operates as a closed cycle system with no loss of coolant as long as power is maintained to the refrigerator. Based on the manufacturer’s specifications and applied load the required maintenance interval is expected to be 4 years. This will further reduce man-made disturbances for helium transfers and cryocooler maintenance and will allow operation in remote areas.

VII. Acknowledgment of Support

The research done for this paper at GWR Instruments, Inc. was supported by the U.S. Department of Energy under Grant No. DE-FG03-95ER81979 and by a commercial contract with the Bundesamt fuer Kartographie und Geodaesie, Frankfurt, Germany.

VIII References

Crossley, D., O. Jensen, J. Hinderer, Effective barometric admittance and gravity residuals, Phys. Earth Planet. Int., 90, 221-241, 1995.

Durcarme, B., P. Melchior, Fine structure of the tidal spectrum eight lunar nodal waves derived from the 15 years series of observation with the superconducting gravimeter GWR T3 in Brussels, Proc. 13th Int. Symp. Earth Tides, Brussels, 1998, in press.

Elstner, C., M. Harnisch, W. Schwahn, Planetary waves in the spectra of air pressure and gravity variations, Proc. Symp. 112, Geod. and Phys. of the Earth Potsdam 1993, (eds. Montag, Reigber), 194-199, 1993.

Goodkind, J., C. Young, Gravity and hydrology at Kilauea volcano, the geysers and Miami, Les Cahiers du Centre European de Geodynamique et Seismologie, 3, Luxembourg, 163-168, 1990.

Harnisch, M., G. Harnisch, Longtime behaviour of superconducting gravimeters derived from observation time series, Marèes Terrestres Bull. Inf., 127, Brussels, 9796-9805, 1997.

Klopping, F., G. Peter, K. Bertis, W. Carter, J. Goodkind, B. Richter, Analysis of two 525 day long sets obtained with two side-by-side, simultaneously recording superconducting gravimeters at Richmond, Florida, USA, Proc. of Second Workshop: Non-tidal gravity changes, Les Cahiers du Centre European de Geodynamique et Seismologie, 11, Luxembourg, 57-69, 1995.

Peter G., F. Klopping, K. Bertis, Observing and modeling gravity changes caused by soil moisture and griundwater table variations with superconducting gravimeters in Richmond / Florida, U.S.A., Proc. of Second Workshop: Non-tidal gravity changes, Les Cahiers du Centre European de Geodynamique et Seismologie, 11, Luxembourg, 147-160, 1995.

Richter, B., The long-period tides in the Earth tide spectrum, Proc. IAG Hamburg, 1, Dept. Geod. Sci. Surv., OSU, Ohio, 204-216, 1983.

Richter, B., Das supraleitende Gravimeter, Deut. Geod. Komm., C, 329, Frankfurt / M., 1987.

Van Dam, T., O. Francis, Two years of continuous measurements of tidal and nontidal variations of gravity in Boulder, Colorado, GRL, 25, 3, 393-396, 1998.

Warburton, R., E. Brinton, recent developments in GWR Instruments’s superconducting gravimeters, Proc. of Second Workshop: Non-tidal gravity changes, Les Cahiers du Centre European de Geodynamique et Seismologie, 11, Luxembourg, 23-56, 1995.