IceMole probe

A specialized probe, called the ‘IceMole’, has been developed as a carrier system for all subsystems to accomplish the Blood Falls mission. The IceMole is a manoeuvrable in-ice probe, which combines a melting head with an ice screw for forward thrust and movement even against gravity (Reference DachwaldDachwald and others, 2014). Partial heating of the IceMole melting head allows navigation of curved trajectories through the ice.

Attached to the melting head is a square pipe with dimensions of 15 cm × 15 cm × 180 cm and sealed by a back plate. The 2.9 kW power for forward melting leads to a maximal velocity of the IceMole of ∼1.1 m h⁻¹. Due to the length of the probe and additional side melters, a minimal curve radius of ∼10 m can be achieved. The meltwater is pumped out of the melting channel to allow the IceMole to be recovered. The probe is retrieved backwards with a steel cable once the trajectory has been completed and a sample taken. The current design allows trajectories with a maximum length of ∼80 m. To retrieve a clean sample from the subglacial brine, the clean access strategy is a multi-step programme (Reference DachwaldDachwald and others, 2014). Just prior to taking the sample, an in situ decontamination procedure is performed. Then the IceMole taps the fissure with a pneumatic needle. In total, 1 L of brine is pumped through the needle into sampling bags located within the IceMole payload bay.

Navigation system

The manoeuvrability of the melting probe requires knowledge about its position and heading. To avoid obstacles and to find the point of interest, the forefield of the probe needs to be scanned. The IceMole has dedicated subsystems to provide this information. One subsystem is responsible for the inertial navigation and guidance (Reference NiedermeierNiedermeier and others, 2014). It uses an inertial measurement unit, which is equipped with accelerometers and gyroscopes. The heading information is provided by a differential magnetometer system with one flux-gate magnetometer integrated into the probe, and a second reference magnetometer on the surface. An axial feed sensor is connected to the screw as an odometer and provides measurement of the distance covered. To provide information on the forefield, the probe is equipped with an acoustic reconnaissance system. This subsystem uses ultrasonic phased arrays integrated in the IceMole head for sonographic imaging. The absolute position is provided by the acoustic positioning system (APS) described in this paper. All sensor information is collected, combined and visualized at the surface.

Principle and overview

The APS provides the absolute position of the probe based on the idea of trilateration (Fig. 2). Trilateration is a method to determine the position of an object by measuring the distances to at least three other objects at known positions. In this project the unknown position is the location of the IceMole, while the positions of the emitters at the surface are known. The distances are determined by measuring the propagation time of the signals from the transducers to the receivers located in the IceMole head. The transducers require high-voltage pulses to achieve a reasonable power output. To avoid high voltages and electromagnetic noise within the probe itself, the transducers are installed at the glacier surface, while the receivers are integrated into the probe’s head. This does not limit the size of the transducers, which need to be ∼10 cm for a resonance in this low-frequency ultrasonic range. To calculate the distance, a precise estimation of the speed of sound in ice is needed. Its value can be determined by measuring the time a signal needs to travel a known distance. This can be done by taking measurements from one transducer to another. Three signal paths are the minimum required for the trilateration. By over-constraining the measurement with more signal paths, the precision of the positioning can be enhanced and the speed of sound can be determined as an additional parameter.

Fig. 2.

Principle of the APS. The absolute position of the IceMole is determined by trilateration of ultrasonic pulses from several emitters.

Emitter and surface components

The transducers (ITC-1001, International Transducer Cooperation) are spherical sonar pulse emitters with a diameter of ∼10 cm and a resonance frequency of ∼18 kHz. They are placed ∼0.5 m below the ice surface for optimal coupling and to reduce the effects of inclusions, smaller cracks and inhomogeneities in the upper layers of the ice. The holes, in which the transducers are placed, are filled with water for good coupling to the glacier ice. Depending on the glacier temperature the holes either stay water-filled or refreeze. The transducer system consists of six stations which are connected and controlled through a field-programmable gate array (FPGA)-based central transducer unit (CTU). The CTU distributes a differential pulse per second (PPS) signal by an electrical connection to the IceMole to synchronize with the other subsystems with an accuracy <1 μs. To distribute the global time, a GPS antenna can be connected to the CTU, otherwise a local reference time is sent. For communication the CTU is connected via 100 Mbit s⁻¹ Ethernet to the local area network of the IceMole. Each station contains one front-end transducer unit (FTU) and one acoustic transducer. The FTU amplifies the signal coming from the CTU and transfers it to the transducer. In addition to signal emission for the APS, the transducers can also be used as receivers. This allows measurements from one transducer to another to be performed in order to investigate the ice properties, especially the speed of sound. Received signals are amplified by the FTU and transferred back to the CTU, where the signals are distributed to a National Instruments USB-6366 data-acquisition module. This module records the signals and is controlled by an operators’ notebook computer. To calculate the position of the probe via trilateration, it is essential to know the position of all transducers accurately. This measurement is performed with a differential GPS receiver and a GPS reference station with a typical accuracy of a few centimetres.

Fig. 3.

Scheme of the APS data-acquisition microcontroller.

Receiver and front-end electronics

Four acoustic sensors in the IceMole head receive the signals from the transducers at the ice surface. Each acoustic sensor is based on a piezoelectric ceramic disc (PIC-255, Physik Instrumente GmbH & Co. KG), which is encased in a metal housing. The material of the disc is lead zirconate titanate (PZT), the disc diameter is 16 mm and its thickness is 3 mm. The housing protects the sensor mechanically and reduces pick-up of electromagnetic noise. The disc is pressed against the inner side of the IceMole head to achieve good acoustic coupling. This also couples one electrical contact of the PZT disc to the electrical potential of the IceMole chassis. The other electrical contact is directly coupled to the front-end electronics. The zero potential of the front-end circuit is also coupled to the IceMole chassis to avoid electrical noise pick-up. Two stages, one with low (+53 dB) gain and one with high (+80 dB) gain, amplify the signals in a frequency range 10–300 kHz. Both stages are digitized via analogue/digital converters with a 2 MS s⁻¹ sampling rate and 12 bit resolution. The digital, differential signals are transferred into the ‘acoustic box’ inside the IceMole payload bay for further handling.

Data-acquisition system

Data acquisition and preprocessing is via the ‘acoustic data-acquisition board’ and the ‘acoustic embedded PC’. These are located inside the acoustic box situated in the payload bay of the IceMole. The FPGA-based acoustic data-acquisition board simultaneously receives the digital signals from all sensors, stores the waveforms and then transfers these via USB 2.0 to the embedded PC. The use of a low-cost, commercially available embedded PC (Raspberry Pi Model B Rev. 2, Raspberry Pi Foundation) is sufficient. The synchronization between the CTU and the acoustic data-acquisition system in the IceMole is via a dedicated electrical signal cable. The CTU sends 1 PPS as a differential signal which is decoupled in the acoustic box. The acoustic data-acquisition board receives this signal and uses it to synchronize the data acquisition. All other communication between the different systems is performed digitally via Ethernet. The data-acquisition system is powered by the 24 V IceMole subsystem power supply. In the acoustic box the power supply connections are galvanically isolated from the IceMole on a separated power converter board.

Measuring protocol and data handling

The operator initiates a measurement cycle via a network command from the surface PC to the acoustic box embedded PC. At the same time the IceMole is halted by shutting down the heaters. By keeping the torque on the ice screw, the IceMole head is pulled against the ice to achieve good acoustic coupling. Then the embedded PC sends another network command to the CTU to start the signal generation. From the start time of the transducer cycle, the embedded PC calculates the start times for the individual data-acquisition windows with respect to the last synchronization signal. By default the sampling window is 32 ms long and starts 1 ms before the signal generation. Several pulses are emitted, separated by 50 ms. In each measurement cycle, 32 pulses are recorded for each transducer as digitized waveforms. After this the data preprocessing is performed. For each sensor all recorded waveforms are averaged and filtered with a Fourier filter to reduce noise, using a passband of 17.3–26.7 kHz. The corresponding window function is defined as 1 for the range 18–26 kHz with Gaussian tails with a 1 kHz standard deviation. From these filtered waveforms the signal arrival times are determined by searching for the first time the noise level is surpassed by a factor of five. The noise level is defined as the standard deviation of the first millisecond of data, which is recorded before the signal generation started. The raw data waveforms and the signal arrival times are transmitted to the central database at the surface. With the determined signal arrival times, speed of sound and known transducer positions the absolute position of the probe is calculated via trilateration.

Figure 4 shows a flow diagram for the entire APS.

Fig. 4.

Flow chart of the APS, showing all essential parts of the system. CTU: central transducer unit; FTU: front-end transducer unit; emPC: embedded PC; DAQ: data-acquisition system.

Proof-of-principle campaign

The principle of acoustic positioning in glacier ice was tested first on Morteratschgletscher in summer 2012. Several holes were melted vertically into the glacier, and they remained water-filled. One ITC-1001 transducer for signal generation and an array of four analogue ultrasonic sensors were inserted into the holes. Several sine bursts, each with five cycles and frequencies of 30, 45 and 60 kHz, were generated. The signals were acquired by a National Instrument USB-6366 data-acquisition module. Several measurements were taken with propagation distances up to 30 m and a depth of several metres. Signals up to 30 m were recorded as shown in Figure 5.

Fig. 5.

Waveforms of different measurements, and schema of the test site of the APS proof of principle. The schema shows the top view of the melted holes with the corresponding distances between the holes. The waveforms were recorded while the receivers were in hole 0 and the transducer in hole 2, 5 or 6. Note the different scales on the y-axis.

System tests in water

Testing the EnEx acoustic navigation concept in the laboratory is a challenge. The production of sufficiently large amounts of ice is time-consuming and expensive. Furthermore, changing the position of the tested equipment is impossible. The substitution of ice with liquid water allows swimming pools to be used, where the equipment can be easily accessible and movable. Since our system works only with P-waves, the main difference is the speed of sound, which is ∼1500 m s⁻¹ in water and ∼3900 m s⁻¹ in ice (IceCube Collaboration, 2011). Besides this difference the acoustic properties of ice and water are sufficiently comparable for system tests.

A local public swimming pool was used, with dimensions 14 m × 17.5 m and a depth of 4.5 m. Here several development stages from the first APS prototype sensor with a laboratory data-acquisition system to the fully integrated IceMole were tested. Figure 6 shows the scheme of the installation for the full-system test in water. The IceMole and the transducers were mounted 2 m below mobile rafts. The rafts were fixed by tethering ropes to the sides of the pool. White cardboard cylinders were mounted on top of the rafts as targets for measuring the distance to reference points with a laser odometer. The position of the rafts on the water was determined from the distance to the reference points.

Fig. 6.

Sketch of the set-up for the full-system tests in water. IceMole is scaled by a factor of 2.

In a first full-system test, the angular sensitivity of the IceMole was measured. The IceMole was located in the centre of the pool and one transducer was moved in several steps around the IceMole with distances of ∼4–8 m. At every step a signal from the transducer was emitted and recorded with the sensors in the IceMole head. The 1/r-distance dependence of the measured amplitude was corrected by multiplying by the distance. The differences between measured amplitude and maximum amplitude are plotted against the angle in Figure 7. The maximal sensitivity is found in the direction of the screw, the minimal sensitivity in the direction of the back plate. To the sides the received amplitude differs from maximal amplitude by <5 dB up to ±140° polar angle.

Fig. 7.

Angular sensitivity pattern of the IceMole measured during a full-system test in water. 0° is in front of the IceMole.

In another test scenario the IceMole was positioned at different locations along a trajectory inside the pool. This test showed the functionality of the positioning system and gave an estimate of the expected uncertainties. For this test a full measurement cycle with four installed emitters was performed at every location. Afterwards the position of the IceMole was reconstructed.

Figure 8 shows the reconstruction of five positions of the raft and the reconstructed IceMole positions along the trajectory inside the pool. The difference between the reconstruction and the expected position is <0.5 m. The constant offset between the reconstruction and the expected position may arise from a systematic shift between the IceMole head and the reference targets on the rafts. The reconstructed points are in agreement with the actual position within the measured uncertainties and are below the required accuracy of 1.0 m. This test showed that the APS was fully functional and ready to be tested in glacier ice.

Fig. 8.

Reconstructed IceMole positions (circles with errors) in comparison with the position of the raft (diamonds). The errors represent the uncertainties of the trilateration algorithm.

First in situ results in glacier ice

The fully integrated APS was successfully tested on Morteratschgletscher in summer 2013 and on Canada Glacier in autumn 2013. The trajectories were driven in the upper layers of the ice within depths of ∼5 m. The speed of sound in the ice (∼3600 ms⁻¹) was determined by transducer-to-transducer measurements. An example of one of the transducer signals received at the IceMole can be seen in Figure 9. The pulse can easily be identified. The reconstructed trajectory of the rover is shown as crosses in Figure 10. Only the signals from transducers with a sufficient signal-to-noise ratio were used for the trajectory reconstruction. For this reason most of the reconstructed positions are based on signals from only three transducers. Due to this and the inhomogeneities in the speed of sound, the errors of the positioning in the Morteratsch field test were 0.3−2 m. The errors represent the uncertainties of the trilateration algorithm. Signals from distances up to 20 m were registered. The first test in Antarctica, in autumn 2013, was also successful. Small cracks, which would be water-filled in a temperate glacier, were identified as a possible critical issue for acoustic signal propagation in a dry glacier. This test showed that acoustic signals are not substantially degraded by this effect and that the quality is primarily determined by the coupling of the acoustic transducers to the ice.

Fig. 9.

Signal from one transducer recorded by a receiver within the IceMole head during the field test in summer 2013 on Morteratschgletscher. The horizontal dashed lines mark the signal detection thresholds. The arrival time was determined to 3.7 ms, which corresponds to a distance of ∼13 m using a measured speed of sound of 3600 m s⁻¹.

Fig. 10.

Reconstructed 20 m long trajectory recorded during the field test on Morteratschgletscher in summer 2013. The dashed line is a crevasse.

To detect very weak signals, and therefore achieve a larger detection range, the frequency filters were optimized. A design study verifies that an infinite impulse response filter (Reference MeyerMeyer, 2011) with a passband of 17–25 kHz and a Chebyshev type 1 filter design (Reference MeyerMeyer, 2011) gave the best results. An increase in the signal-to-noise ratio of about a factor of two was accomplished. Improved front-end electronics with an optimized two-stage analogue filter were developed for the summer 2014 field test. Further analysis shows that the accuracy of the positioning can be further increased by using more transducer signals, a higher signal power and by improving the coupling of the receivers and transducers to the ice.