A Martian acoustic anemometer

The turbulent eddies that are found just above the interface between a planetary surface and its atmosphere, are the dominant mechanism by which heat, momentum, and molecules are transferred between the surface and the atmosphere (e.g., Priestley, 1959). On Mars, we have yet to fully quantify and understand these processes, how they mirror those on Earth, and how they differ. They represent one of the most open areas for study in Martian atmospheric dynamics. These boundary layer processes control all of the Aeolian effects at the surface (which likely now dominate surface modification on Mars) (e.g., Greeley and Iverson, 1985), as well as the stability of volatiles (e.g., water) at and near the Mars surface (e.g., Chittenden et al., 2008). While the small heat capacity of the atmosphere means it is not a strong forcing on the surface itself, the heat transferred from surface to atmosphere has a controlling effect on the diurnally varying height of the atmospheric convective boundary layer (e.g., Petrosyan et al., 2011). This in turn directly controls the size of dust devils (Renno et al., 1998), and likely other dynamic processes in the Mars atmosphere.

The strong vertical wind shear that develops at the surface—atmosphere interface typically overwhelms the stable stratification in the atmosphere, even on Mars where the stability can be quite strong at night, and turbulent eddies result (Petrosyan et al., 2011). These turbulent eddies actually carry the fluxes of heat, momentum, and molecules through the overturning boundary layer, working to reduce the vertical wind shear, or reduce temperature or molecular constituent (e.g., humidity) gradients in the boundary layer. The eddy diffusion that they perform is typically much larger than the still air diffusion of heat, momentum or molecules through an atmosphere (Preistley, 1959). By resolving the turbulent eddies with a fast-response, sensitive three-dimensional (3-D) wind sensor, one can directly measure the fluxes of heat, momentum, and molecules between the surface and atmosphere (Swinbank, 1951). This technique is widely used in terrestrial atmospheric boundary layer studies and is a key approach in understanding Earth's surface—atmosphere interaction. The key to enabling such a direct measurement of these effects is a wind sensor that can resolve not only the horizontal winds, but also the generally much more gentle vertical winds and the horizontal wind perturbations associated with the turbulent eddies. These fluctuating winds can have short periods with small amplitudes, both dependent on the altitude at which one measures them. Close to the surface (e.g., 1.5 m altitude), measuring at 10–20 Hz and with a sensitivity of ∼5 cm/s is required to capture the dominant components of the turbulent eddy spectrum (McBean, 1972).

None of the wind sensors that have been (or are currently planned to be) sent to Mars could adequately capture the behavior of the turbulent boundary layer. The sensors that have been flown were based on either wind deflection (with very coarse results) or a hot-film/hot-wire approach, where the wind speed is proportional to the heat lost from a heated element. However, in the low density Martian atmosphere (about 1% that of terrestrial sea-level conditions), the advective heat loss is relatively small, and radiative heat losses are significant in comparison, leading to difficulties in calibrating such an instrument. Even more problematic are the limits on sensitivity and response time that this measurement approach yields, with most of the sensors only able to measure winds above 1 m/s and having a response time of about 1 s (e.g., Gomez-Elvira et al., 2012).

We have meteorological samples (including horizontal winds) from five sites on Mars (two Viking Landers, the Pathfinder Lander, the Phoenix Lander, and from the MSL rover locations as it moves). The Phoenix Lander's wind sensor, a telltale, was sampled even less frequently than the others since it required photographing it to return data (Petrosyan et al., 2011). The MSL wind sensors can theoretically sense 3-D winds, but they are poorly accommodated on the rover and are also partially broken, making their results less useful than planned (Gomez-Elvira et al., 2014). The Pathfinder wind sensor was never fully calibrated (Schofield et al., 1997). The best wind measurements on Mars to date have been those from the Viking Landers in the late 1970s.

To achieve the scientific goals of measuring the fluxes of heat, momentum, and molecules through the atmospheric boundary layer, we need to send to Mars a wind sensor with a faster response time, and with more sensitivity than those already sent to Mars. Terrestrially, turbulent eddies in the atmospheric boundary layer are regularly measured using the technique of sonic anemometry, which is the gold standard for atmospheric wind measurements on Earth. We have developed such an instrument for Mars, and will discuss the challenges of adapting the terrestrial instrument for similar performance under the much more challenging Martian conditions.

Sonic anemometers have been in scientific use to monitor terrestrial boundary layer flows since the 1960s (Kaimal and Businger, 1963). The technique essentially measures the acoustic travel time in opposing directions, where the direction going with the wind will have a slightly faster travel time (see Fig. 1). Three-dimensional wind measurements are built up from measurements along three separate axes. If the molecular makeup of the atmosphere is known, the average of the travel times can also be used to infer the atmospheric temperature via the sound speed. Fundamentally the measurement that must be made is a precise acoustic travel time.

Commercial sonic anemometers typically measure the acoustic travel time using a pulsed signal emitted from a piezo transducer. Facing that emitter is an identical transducer used as a receiver. The pulse travel time is typically measured by a zero crossing after the amplitude exceeds a threshold. In the relatively high signal-to-noise environment that is available using standard piezo transducers in terrestrial conditions, this is sufficient to achieve measurement repetition rates of >20 Hz and sensitivities of <5 cm/s. This meets the science goals for resolving the bulk of the energy in the turbulent eddy spectrum in the atmospheric boundary layer. Because the sonic anemometer can also yield air temperature (if the atmospheric molecular makeup is known), the sonic anemometer alone can be used to produce eddy fluxes of both heat and momentum. Combined with a fast response molecular sensor, fluxes of various volatiles (e.g., water or methane) between surface and atmosphere can also readily be measured using sonic anemometry on Earth.

The next generation of scientific questions to be pursued about Mars atmospheric boundary layer, and surface atmosphere exchange will require the same type of turbulent eddy measurements as are commonly performed on Earth using sonic anemometers. This technique is the best choice for Earth, and we believe it is also the best choice for Mars, but it is not without challenges. We will document the difficulties of adapting the terrestrial measurement approach to Mars here.

To measure the acoustic travel time, one must couple acoustic energy from a transducer, into the Martian air, across a reasonable path length in the air, and then back into another transducer at the other end for reception. There are significant losses in coupling between the transducers and the Martian air because of the acoustic impedance mismatch. The density of Mars air is typically only about 1% that of Earth air (Mars surface pressure is about 6 mbar and is dominated by CO₂), and the sound speed is about 2/3 that on Earth. This results in an acoustic impedance that is reduced from Earth air by more than 2 orders of magnitude.

This impedance mismatch has always been a challenge in air-coupled ultrasonics in terrestrial conditions (e.g., Schindel et al., 1992). While piezo transducers can be driven with high energy signals, they are already relatively poor at coupling to Earth air because of their high intrinsic acoustic impedance. Reducing the medium's acoustic impedance by 2 more orders of magnitude makes the problem significantly worse (see Fig. 2). Matching layers (e.g., Alvarez-Arenas, 2004) have been used in air-coupled ultrasound to better couple piezo transducers to air, in some cases with good success. Alternatively, transducers with very lightweight membranes have also been employed, driven by electrostatic forces or alternatively by piezo materials (e.g, Hutchins et al., 1997). In both cases, successful coupling with the low acoustic impedance Earth air has been achieved, enabling air-coupled ultrasound. More recently, fully micromachined capacitive and piezo ultrasonic transducers have been developed, both of which also show good coupling with Earth air (e.g., Chimenti, 2014).

We chose to use a capacitive transducer, with its intrinsic low acoustic impedance to optimize our coupling to Mars extremely low acoustic impedance air. A search was conducted among the state of the art of capacitive transducers in the early 2000s and the best performance was found using transducers from MicroAcoustic Instruments, Inc. While the two-way transduction signal was dramatically reduced using Mars air compared to their performance in Earth air, there was still sufficient signal to enable the instrument to overcome the high penalty coupling with the low acoustic impedance Mars air.

In addition to the acoustic impedance mismatch transduction losses, the low density CO₂ atmosphere of Mars also has relatively high attenuation when compared to Earth air (Williams, 2001). This is particularly true for higher frequencies, where the large mean free path on Mars starts to become similar to the wavelength of the sound waves, greatly enhancing the attenuation they experience. Consequently, for a typical path length between transducers of 15 cm, only frequencies below about 100 kHz are practical for use on Mars (see Fig. 3).

The challenges in getting signal from transducer to air, through the air and then back into a receiving transducer cannot be solved on Mars simply by adding more power to the instrument. Mars landers and rovers are typically operated on power budgets of roughly 100 W, where the majority of that is used in transmitting data to Earth (or more commonly orbital spacecraft overhead), survival heaters, or the main spacecraft flight computer. This leaves a budget of only a few Watts for the typical instrument. This is especially true for a meteorological instrument that is presumably designed to operate continuously, so the power draw is a constant, unlike other instrument types (e.g., cameras) that may only operate for a very small fraction of the entire day. Our instrument design in a flight configuration would draw roughly 7 W, and yet there are strong incentives to reduce that significantly.

The mass of an instrument is also a precious commodity on a landed spacecraft on Mars. Decelerating to safely land at the surface of Mars is easier for lighter spacecraft. Typical instruments at the surface of Mars are measured in single kilograms. Our sonic anemometer in flight configuration would be just over 1 kg, including a 70 cm boom to place the sensor further away from the lander to achieve a better measurement.

The 70 cm boom that we included in our flight design instrument was a trade-off between instrument mass, and the lack of interference in the wind measurement from the perturbations of the winds from the lander itself. Typically on Earth, wind measurements are taken from towers several meters above the ground, but especially separated from nearby obstacles by 10 times the height of the nearest obstacle. This is not feasible on Mars without paying a tremendous penalty in the mass of the instrument to deploy away from the lander and with a significant tower. Instead, if the sensor is placed away from the body of the lander a distance roughly equal to the size of the lander, the wind perturbations can be minimized with a minimum of mass and complexity to separate the sensor from the lander (Lenoir et al., 2011).

To the best of our knowledge, the first known testing of a sonic anemometer under Martian conditions was led by Robert Haberle at NASA Ames in the mid 1990s. They tested a commercial terrestrial sonic anemometer in the NASA Ames Mars Wind Tunnel (MARSWIT) and found it performed well as the pressure dropped until reaching a pressure of about 100 mbar, at which point the instrument no longer produced usable results (Haberle, 2016). Towner et al. (2000) and Wilson (2003) discussed an unconventional type of 2-D resonant sonic anemometer adapted from a commercial instrument by FT Technologies that was originally intended to be the wind sensor for the Beagle Lander. However, development stopped on that instrument before flight and an alternative hot-film style instrument was flown. Reportedly, the terrestrial instrument on which this was based worked down to ∼15 mbar and with modifications down to ∼4 mbar (Wilson, 2003), though experimental proof of the latter was apparently never provided in the scientific literature. Such an unconventional design in any case was intrinsically only 2-D, was not of an open sensing volume design, had admitted problems with accurately measuring temperature, and would be ill-suited to detect and measure higher frequency eddies of great interest in the Martian boundary layer.

Our own instrument development started in 2002 with NASA PIDDP funding (e.g., Banfield et al., 2003; Banfield et al., 2004; Dissly et al., 2005; Banfield et al., 2006; Banfield et al., 2007; Banfield et al., 2012; Banfield and Dissly, 2012; Rafkin et al., 2012; Hudson et al., 2012; Banfield, 2012; Rafkin et al., 2013; Rafkin et al., 2014). One of the three groups (U. Warwick) that we funded to produce the candidate transducers we tested went on to publish the results of their independent tests on their transducers (Davis et al., 2005; Davis et al., 2007). Wilson eventually teamed with the Warwick group to further explore developing a sonic anemometer for Mars (Wilson et al., 2008). They developed custom transducers similar to the early designs of MicroAcoustic Instruments and at several conferences reported their performance under conditions approaching (but still well above) those on Mars (Leonard-Pugh et al., 2011, 2012). We believe that this is the extent of work on developing sonic anemometry for Mars, and that we have produced the first working 3-D open sensing volume sonic anemometer for Mars in this work.

As mentioned above, we selected MicroAcoustic Instruments BAT^™ transducer technology as the best available to couple with Martian air at the time of our search in 2003. This decision was made following an experimental head-to-head comparison in rarefied CO₂ of the best candidate sensors available at the time, where the BAT transducers grossly outperformed the others (in fact the others would not have been able to form a functioning instrument). Working with MicroAcoustic Instruments, some of their standard commercial BAT-1 transducers were then customized over the years for our application. In particular, MicroAcoustic Instruments directed significant effort toward: (i) reducing the diameter of the transducer housings; (ii) reducing their bias voltage requirements; (iii) increasing their sensitivity to the frequencies of interest on Mars; and (iv) ensuring both survivability and solid performance under what are fairly rigorous diurnal temperature variations on Mars. The resulting customized, proprietary MicroAcoustic mini-BAT^™ sensors can be seen in the various photographs of this report.

The reduction of sensor head diameter was important for this application. The standard BAT-1 transducers from MicroAcoustic (which are packaged for use in the field of non-destructive testing) are 3.4 cm in diameter with an active diameter of 1 cm. For an acoustic path length of only 15 cm in a Mars-bound anemometer such large transducer heads presented a significant obstacle in the wind, biasing wind measurements for a large range of angles on either side of the flow directly aligned along the transducer direction. This is a problem for sonic anemometers on Earth as well, where it is mitigated by minimizing the transducer size relative to the inter-transducer spacing, and then calibrating the detailed wind speed response of the instrument to all wind directions (e.g., Wyngaard and Zhang, 1985). The customized mini-BAT transducers employed here were considerably reduced in size, being only 1.5 cm in diameter for an active area of 1 cm diameter. This still resulted in a larger aspect ratio (transducer diameter/inter-transducer spacing ∼1/10) than is typical in terrestrial systems, but was required due to the low signal-to-noise environment of Mars. Increasing the path length was not possible due to the significant attenuation in Mars air, and decreasing the transducer head size was not possible due to the low headroom of the signal above the noise. Reducing the active area of the transducers would reduce the signal still further, while not likely changing the noise appreciably (see Fig. 4). We found that we could successfully calibrate out these wind shadow effects from the instrument, even with the relatively large transducer size/spacing aspect ratio that remained.

Ideally, we would be able to use each transducer as both a transmitter and a receiver. In principle, BAT transducers can be used this way. However, in practice we did not do this, as the need to switch quickly from transmit to receive using a duplexer was expected to introduce significantly more noise into our system, and thus likely not allow the instrument to meet the science goals outlined above. Instead, we chose to use two transducers on each side of each axis. One transducer of the pair was dedicated to emitting the acoustic signal, while the other would receive. We would emit sound from both sides of an axis at the same time, and receive the acoustic signal roughly 500 μs later on the dedicated receive transducers.

The typical electrical power on a spacecraft is delivered at roughly 28 V, fluctuating by +/−∼4 V as the spacecraft batteries charge and discharge. The standard BAT transducers from MicroAcoustic Instruments use relatively high voltage biases (∼200 V) to pull the membrane to the backplate, and allow modulations of the voltage to produce roughly linear motions of the membrane. Producing these high bias voltages, and especially the related high drive voltages to oscillate the transducer membrane meant that we would need special power handling electronics and more complex amplifiers. This meant increases in the instrument mass, volume and power that we had incentive to minimize. MicroAcoustic Instruments therefore redesigned the custom mini-BATs so that their optimal performance would occur instead at a reduced operating voltage of 80 V (and further reductions are possible in the future). While this was not yet as low as our goal of <28 V, it did mean simpler, smaller and lower power electronics than we would have had with the higher bias and drive voltages of the standard transducers. This redesign of the mini-BAT sensors for optimal performance at lower voltages had an added bonus: both transducer sensitivity and the signal to noise ratio increased slightly, improving the performance of the instrument as a whole in addition to reducing its resource footprint.

Since the finalization of this prototype instrument, further improvements to the transducers have been made. In particular, newer designs of mini-BAT have recently been created and tested at MicroAcoustic Instruments which have operating bias voltages of only 32 V and which produce ∼4× the signal amplitude for a source-receiver pair in 6 mbar CO₂. These new transducers would allow us to reduce drive voltages (and thus instrument power) and/or transducer sizes (and accompanying wind shadowing) through a move to a single transducer housing on either end of each axis.

Surface temperatures on Mars are somewhat more extreme than here on Earth. Typical temperatures for tropical sites on Mars fluctuate between ∼0 °C and ∼−80 °C (depending upon where on the planet you are) with extremes approaching ∼+20 °C and ∼−132 °C (i.e., the frost temperature for CO₂ at the surface pressure of 6 mbar), biased toward the colder temperature as one moves away from the tropics. As a result, any exposed sensors outside the protected interior of a rover must continue to perform their duties under significant daily temperature fluctuations. Throughout design iterations of the customized mini-BAT sensors, MicroAcoustic Instruments, therefore, subjected operating sensors to the lower extremes of temperature expected on Mars in order to try to ensure their survivability and to help in evolving the various designs. In particular, any one sensor design of interest was typically exposed to 90 sinusoidal temperature cycles between +20 °C and −60 °C in a dry air environment followed by ∼3 sinusoidal temperature cycles down to as low as −135 °C in a dry nitrogen environment. The former fluctuations were created using an environmental chamber programmed for the purpose, while the latter used a suitably-programmed liquid-nitrogen fueled environmental chamber. Each of the sensors employed in the various studies of this report went through such temperature cycling prior to their use. One strength of the developed mini-BAT transducers for the present application is that their sensitivities were found to actually rise somewhat with falling temperature, such that any acoustic anemometer developed around them will perform better as the temperature falls on Mars. Though we had wished to run much longer duration temperature studies (of 600 cycles or more), we have not done so to date. Nonetheless, we have not seen anything experimentally that suggests these mini-BAT transducers cannot be made to survive and perform admirably at Mars temperatures on extended missions.

In terms of mini-BAT sensitivity, there are no calibrated receive sensors available (as far as we are aware) capable of measuring absolute pressure levels of ultrasound in CO₂ at low pressure. For this reason it was not possible to determine absolute pressure levels of developing prototype mini-BATs as the development unfolded. Instead, to gauge our progress along the way, we chose to define a more practical “headroom factor” specific to the Mars acoustic anemometer development. We began with a computer model for the operation of the instrument as a whole. This model included full three-axis operation in a 6 mbar CO₂ environment and allowed us to change wind speed and direction and to input synthesized receive signals that closely matched real receive signals in terms of signal shape, amplitude and noise. By then reducing the signal amplitude of the receive sensor signals within the model (keeping all other factors the same) we were able to determine the point at which the instrument stopped meeting its science goals for wind speed accuracy of 5 cm/s at an update rate of >20 Hz. The point at which this occurred then gave us the minimum acceptable signal amplitude for any one source-receiver pair within an instrument that was just meeting its overall science goals. Any subsequent real measurement of pitch−catch signal amplitude for a real source-receiver pair operating in 6 mbar CO₂ could then be divided by the minimum acceptable signal amplitude value to provide our “headroom factor.” As such, the headroom factor is then a simple multiplier that shows us just how many times over the minimum acceptable signal level we are with any one developing prototype sensor pair. For the mini-BAT sensors employed within this report the headroom factor at room temperature was approximately ∼17×. Though this may seem like considerably more headroom than is necessary, appreciate that various (and often numerous) compromises are imposed upon instrument developers through the process of accommodation on an actual rover destined for flight, such that a starting headroom value of 17× might be whittled down through circumstances beyond the developers' control. Also, there are unknowns that one cannot fully plan for, such as an imposed close proximity to other instruments and rover systems during rover integration that can lead to a noise environment (electromagnetic, electrical, acoustic, etc.) that is not always fully known in advance. As a result, we feel comfortable that our current headroom value of ∼17× is sufficient to meet the science goals for an actual rover mission to Mars but we would prefer that it not go much lower than this.

While the Microacoustic Instruments transducers couple to terrestrial sea level air very well with extremely broad bandwidth, unfortunately, in spite of the improvements that were made in customizing them for Mars performance, the headroom of signal over noise under Martian conditions and a 15 cm propagation distance were not great. This meant that we would not be able to achieve our science goals using the simple techniques of measuring acoustic travel times that are used in terrestrial sonic anemometers. Instead, we used the technique of pulse compression, which uses the bandwidth of the transducer and an extended chirp duration to achieve a higher signal to noise result, while still producing precise acoustic travel times (Gan et al., 2001). In our case, we would transmit chirps of duration about 400 μs that varied in frequency from 20 kHz to 80−200 kHz linearly over that duration (see Fig. 5). Other options for the transmitted signal are possible, such as binary codes or other methods of varying the frequency, however we restricted our attention to linear chirps for simplicity. To arrive at the final choice of our transmit signal, we tested various combinations of the chirp's duration, and low and high frequency using a numerical model of the instrument to determine its sensitivities. From this exercise, we designed our final prototype instrument to use an un-windowed linear chirp from 20−120 kHz with a duration of 400 μs.

The pulse compression technique estimates the acoustic travel time by correlating the received version of the chirp with a copy of that transmitted. At the lag equal to the actual travel time, the correlation is maximized. This technique is able to extract accurate travel times down to and possibly below signal to noise levels of 1.

The only drawback of this technique is that it is a relatively complex technique. That requires the emitted chirp to be an accurate representation of the idealized chirp that is stored in memory, which places constraints on the fidelity of the production and amplification of the drive signal. It also sets requirements on the computational capabilities of the instrument. From the surface of Mars, it is not feasible to downlink the full measured waveform for each chirp of the six different directions on a 3-D sensor at the desired 20 Hz sampling rate. Instead, the data must be processed to yield acoustic travel times on board the instrument, only downlinking six values (the travel times) for each measurement. This meant that the CPU with which we performed the correlation between each received chirp and the emitted chirp had to have sufficient processing speed to accomplish these computations in real time. This meant more power and mass for the instrument. In our flight configuration instrument, we estimated that the CPU consumed 4 W on its own, the lion's share of our overall power usage.

Our final prototype was designed around a Connect Tech “Freeform PCI-104” FPGA board on a PC-104 form-factor with a Xilinx Virtex-5 FPGA. It incorporated a Xilinx Microblaze soft-core processor with floating point running at 100 MHz. The design had the FPGA generate a digital chirp (14 bit) that was passed to a D2A (5 V), then amplified to +/−80 V and then biased with a DC +80 V before being sent to the transmitting transducer. The overall transmit cycle was 1 ms in duration, with 1024 samples at 1 MHz. The chirp itself was 400 μs at the beginning of the transmit cycle. On the receive side, the signal from the transducer was amplified by a MicroAcoustic Instruments Q-Amp followed by a preamplifier to produce a +/− 0–2.5 V signal. This was then digitized by a 16 bit A2D also at a sampling rate of 1 MHz and a resolution of about 0.05 mV/DN. The noise in the electronics without transducers connected was only about 1 DN, whereas once we connected transducers, the noise level under quiescent conditions in the lab was about 2 mV (40 DN). To process these signals, the FPGA would read and store the digitized signal in a FIFO. The Microblaze software would read the data from the FIFO and perform the pulse compression. This was the brunt of the computing power, requiring two floating point FFTs and one inverse FFT to pulse compress each chirp. We correlated the full 1024 samples from each chirp, but some optimization could have been possible with shorter windows at the expense of a slightly more complex algorithm.

We tested the final prototype (1-D breadboard) version of our instrument (see Fig. 6) on a terrestrial stratospheric balloon, as well as in a dedicated Martian Wind Tunnel at Aarhus University in Denmark (Holstein-Rathlou et al., 2014). The stratospheric balloon test (see Fig. 7) floated the instrument at about 110 000′ altitude where the pressure is about 8 mbar and the temperatures are ∼−40 °C, i.e., very similar to the conditions at the surface of Mars. This test was a good check on the survivability of our instrument in the harsh radiative heating/cooling conditions that one would encounter at the surface of Mars, but it also forced us to design the prototype to autonomously operate and record data as it would have to on Mars. It was less valuable as a test of the performance under controlled wind conditions, as the winds around the balloon gondola were much less steady than expected. Turbulent conditions dominated essentially the whole duration of the flight, even when the balloon was relatively stably floating at the peak target altitudes. Consequently, it was difficult to compare our wind measurements with any calibrated values. We had assumed that we could compare our measurements with the vertical motion of the gondola (determined from a combination of the pressure change and the GPS altitude), but that proved difficult in that the steady vertical winds due to the upward or downward motion of the gondola were a small fraction of the turbulent winds seen around the gondola itself. We expected the balloon to advect horizontally with the wind, and thus have very little horizontal winds evident at the gondola level, and then, when the balloon was descending, to find laminar flows impinging on the wind sensor, which was among the lowest pieces of the balloon stack. However, turbulent wind was found at all times in the flight, perhaps suggesting that vertical shear was occurring everywhere, and while the balloon itself may have been advecting with the horizontal winds, perhaps the gondola (which was ∼100′ lower) may have experienced some horizontal winds. Alternatively, the radiative heating/cooling of the gondola itself may have produced convective flows, exchanging heat with the environment, and producing the strong turbulence we observed. We do not believe our results are indicative of ambient environmental turbulence at these levels, but rather the specific turbulent flow that was induced by our gondola and experienced at the un-optimized location of our sensor in the balloon stack.

The wind tunnel at Aarhus University (see Fig. 8) proved to be an excellent proof and calibration facility for our instrument. In this unique facility, the test chamber can be filled with CO₂. The chamber's pressure can be reduced to ambient Mars surface pressure values (i.e., 4−10 mbar). The air temperature can be controlled down to typical Mars temperatures (i.e., −40 °C), and winds can be produced up to about 15 m/s. We cooled the air in the chamber down to −40 °C to minimize the humidity that would be residual in the vacuum chamber. The attenuation of Mars air is a strong function of the humidity. While we do not expect significant humidity on Mars itself, within a vacuum chamber at low pressure, the majority of the residual gas is often water vapor, slowly desorbing off the chamber walls. To reduce this effect, rather than bake the water off by prolonged heating under very low pressure, we instead cooled the chamber to condense the water vapor onto the cold surface of the cooling coils. Holding them below −40 °C reduced the water vapor mixing ratio in the chamber to well below 0.1%, where its effects on the atmospheric acoustic transmission in the frequencies of interest was negligible.

We installed our 1-D instrument in the wind tunnel with the ability to rotate it around a vertical axis, so that the sensor would be aligned with the wind or to rotate it about 100 deg away from that in one direction and 170 deg away from that in the other. The wind tunnel was set up with flow straighteners upstream of our wind sensor, but no other upstream obstacles. The wind tunnel cross section is large enough that introducing the wind sensor itself reduces the throat cross section of the wind tunnel a negligible amount, so previous calibration of the wind tunnel without our sensor in place was carried over without change to our experiment.

The performance of our instrument was slightly less good than we had expected, we believe because of some extra noise in the system possibly from the electrical feedthroughs into the vacuum chamber between the transducers and their preamplifiers. We found that our instrument worked tremendously well at 10 mbar, able to measure winds with a precision of about 5 cm/s with a repetition rate of 20 Hz (see Figs. 9 and 10). However, when we lowered the atmospheric pressure to 6 mbar, the signal to noise decreased (particularly impacting the one direction that was already experiencing a higher noise level than the other, probably due to a problem with a feed-through to the vacuum chamber) so that only one direction of our 1-D instrument was able to measure realistic acoustic travel times. The performance of the better direction was good enough that had both directions worked this well, we would have still be able to achieve the 5 cm/s at 20 Hz metric of success, even at only 4 mbar (well below typical Mars surface pressure).

The detailed measured wind speed as a function of the azimuth of the wind relative to the inter-transducer direction allowed us to produce a correction algorithm for the wind shadowing due to the finite size of the pair of transducers at each end of the sensing volume. The iterative technique we developed yielded corrections to within 1% of the unperturbed wind speeds, effectively removing the problem of transducer wind shadowing, at least for the 1/10 aspect ratio transducer/spacing that we were using. Smaller transducers would make this correction even easier.

Our measurements in the wind tunnel at 10 mbar were precise enough that even at only 2 m/s winds, we were able to precisely identify that the wind tunnel had an 8 deg cross-flow (not parallel to the wind tunnel axis). This cross-flow was subsequently confirmed using a mechanical wind vane in the tunnel.

We have demonstrated that an acoustic anemometer can be successfully implemented for use in the Martian atmospheric boundary layer. Such an instrument can open new avenues of inquiry into Martian atmospheric science, including directly understanding the forces controlling Aeolian processes on Mars and surface atmosphere exchange of heat and volatiles. We expect that such an instrument will be crucial as we move further in exploring Mars and delivering spacecraft to and from the surface of Mars.

We acknowledge funding from NASA's PIDDP program for this instrument development effort. Considerable technical help was provided by Kelly Kanizay, James Lasnik, Mark Richardson, Ian McEwan, and Jim Waters. We thank Jon Merrison and Jens Jacob Iverson for assistance and the use of the Danish Mars Wind Tunnel. We thank Troy Hudson and also the Columbia Scientific Ballooning Facility for their help in making the stratospheric balloon test flight a success.