THE GALILEO PROBE ATMOSPHERE STRUCTURE INSTRUMENT ALVIN SEIFF San Jose State University Foundation, Department of Meteorology, U.S.A. and T. C. D. KNIGHT Martin Marietta Space Systems, Denver, Colorado, U.S.A. Abstract. The Galileo Probe Atmosphere Structure Instrument will make in-situ measurements of the temperature and pressure profiles of the atmosphere of Jupiter, starting at about 10^-10 bar level, when the Probe enters the upper atmosphere at a velocity of 48 km s^-1 and continuing through its parachute descent to the 16 bar level. The data should make possible a number of inferences relative to atmospheric and cloud physical processes, cloud location and internal state, and dynamics of the atmosphere. For example, atmospheric stability should be defined, from which the convective or stratified nature of the atmosphere at levels surveyed should be determined and characterized, as well as the presence of turbulence and/or gravity waves. Because this is a rare opportunity, sensors have been selected and evaluated with great care, making use of prior experience at Mars and Venus, but with an eye to special problems which could arise in the Jupiter environment. The temperature sensors are similar to those used on Pioneer Venus; pressure sensors are similar to those used in the Atmosphere Structure Experiment during descent of the Viking Landers (and by the Meteorology Experiment after landing on the surface); the accelerometers are a miniaturized version of the Viking accelerometers. The microprocessor controlled experiment electronics serve multiple functions, including the sequencing of experiment operation in three modes and performing some on-board data processing and data compression. 1. Introduction When the Galileo Probe enters the atmosphere of Jupiter in December, 1995, an Atmosphere Structure Instrument (ASI) will make in-situ measurements of the thermal structure of the atmosphere, i.e., the variation of temperature, pressure, and density with altitude. Sensors of atmospheric temperature and pressure, and acceleration of the probe center of gravity will permit a number of key properties of the atmosphere to be derived, as demonstrated by the Pioneer Venus probes (Seiff et al., 1980a). The Galileo experiment will operate over an altitude range of about 870 km in two measurement modes, one to accommodate conditions of high-speed entry at low ambient density; the other, the very different conditions of parachute descent. The entry mode begins at a nominal ambient density threshold of 10^-11 kg m^-3 (presently believed to be about 750 km altitude above the 1 bar level) where the Probe deceleration is expected to be about 15 micro g. From measurements of Probe deceleration under the action of atmospheric drag, atmospheric densities will be derived. The density profile is integrated above a given altitude to define the pressure at that level, and the temperature profile is then obtained through the equation of state, given the variation of atmospheric mean molecular weight with altitude. This mode of operation is continued until the Probe deploys its parachute, nominally at the 100 mb level (47 km above the 1 bar level). Then, during the subsequent parachute descent of the Probe to the nominal end-of-mission depth at 16 bars, the thermal structure of the atmosphere is defined by measurements of temperature, pressure, and acceleration in descent mode. Current estimates are that the mission could extend to below 20 bars, perhaps as deep as 25 bars. At entry, the Probe is effectively a 48 km s^-1 meteor, enveloped by a bow-shock wave and a thin shock layer of ionized, luminescent gases at extreme temperature (~15000 K at peak). Under these conditions, measurements of the ambient atmosphere by means of conventional low-density sensors are clearly infeasible; it is futile to extend sensors into the shock layer or through the bow-shock wave, because they quickly burn away and, outside the probe bow wave, develop shock layers of their own. Measurements of atmospheric density by way of Probe decelerations, however, provides a direct means of sensing the atmosphere. This concept was tested and demonstrated in the Earth's atmosphere in 1973 (Seiff et al.) and has been applied with good results at Mars and Venus (Seiff and Kirk, 1977; Seiff et al., 1980a). After the parachute is deployed, the heat shield is jettisoned and temperature and pressure sensors are exposed to the ambient atmosphere. Acceleration measurements are continued, but at much less frequent intervals, and in the absence of large vertical winds, continue to define atmospheric density. From the three measured state variables, temperature, pressure, and density, the atmospheric mean molecular weight may be determined. Other instruments on the Probe, including the mass spectrometer and the helium abundance detector, will probably determine the molecular weight more accurately, however. In these circumstances, the accelerometer data may be used to define the magnitude of large vertical winds in waves or gusts. Altitudes relative to any convenient reference level (e.g., the 1 bar level) are defined by the temperature and pressure data integrated in the equation of hydrostatic equilibrium. These data will be used to establish the altitudes of measurement for use by all Probe experiments. They also define the rate of descent, which is necessary to the analysis of Doppler wind experiment. In addition, atmospheric turbulence, radius to the center of the planet, and Probe angle of attack are sensed or derived. The nominal end of mission, at 16 bars, is at an altitude of about -120 km below the 1 bar level. The vertical measurement range of about 870 km will be covered in a little over one hour - the initial 700 km in less than 4 min. The time in descent on the parachute at velocities from the order of 100 to 30 m s^-1 could be up to 75 min. The instrument electronics perform a number of essential functions. They receive and execute commands from the Probe systems and thereby set the experiment mode (calibrate, entry, or descent), control the measurement sequences, select sensor ranges, collect data from the three sensor sets, amplify signals, perform A/D conversions, do some onboard data processing, and condition instrument power. The microprocessor controlled electronics were designed and built by Martin Marietta Aerospace, in Denver, Colorado. This is a fourth generation entry probe atmosphere structure instrument. Its predecessors were used on the four probes of the Pioneer Venus Mission (Seiff et al., 1980b), on the Viking Mission to Mars (Seiff, 1976), and on the Planetary Atmosphere Experiments Test (PAET) entry probe into the Earth's atmosphere (Seiff et al., 1973). Best features of earlier instruments were retained and some new capabilities have been added. In this paper, we discuss the design approach briefly, and describe the instrument, both the sensors and the electronics, but refer the reader to earlier publications for more thorough treatment of some topics. Two photographs of the instrument are given in Figure 1. The first shows a close-up of the electronics housing with three pressure sensors mounted on its side; and the second, the complete instrument, with the pressure sensors thermally insulated for installation on the Probe. Instrument components will be discussed below. Fig. 1a. Fig. 1b. Fig. 1. Photographs of the Galileo Probe Atmosphere Structure Instrument. (a) Electronics housing and pressure sensors, with the inlet manifold installed. Cables from the three pressure sensors join in a single connector which mates with one on the electronics housing. (b) The complete instrument, ready for installation on the Probe. The pressure sensors have been thermally insulated. The accelerometers are within the white housing. The temperature sensor is mounted on a block used in laboratory testing. 2. Scientific Objectives and Expected Results Some scientific objectives have been implied above. For convenience, they are more fully summarized here: (1) To accurately define the state properties as a function of altitude below the 100 mb level to ~20 bars. (They have never been measured below the 1 bar level.) (2) To define the currently highly uncertain state properties of the upper atmosphere. (3) To measure the stability of the atmosphere, and identify convective layers and stable layers, where they exist. (4) To detect cloud levels from changes in lapse rate at their boundaries. (5) To define state properties within the clouds, and thus provide supplementary information on cloud composition. (6) To search for and characterize wave structures in the atmosphere. (7) To search for and measure intensity and scale of turbulence in the atmosphere. (8) To measure vertical flow velocities above a threshold of about 0.3 m s^-1. (9) To establish an altitude scale for use in correlating all Probe experiment data. (10) To define the probe vertical velocity, necessary to the analysis of the Doppler wind experiment. One area of high interest to the experiment team is atmospheric dynamics. Key questions concern the nature of the Jovian circulation, and the convective processes (Ingersoll, 1981, 1990). Is the circulation very deep, extending well toward the planet center and driven by internal heat, or is it superficial, driven by energy of condensation of cloud species and/or solar heating, or is it a combination of these phenomena? Are gravity waves present and instrumental in maintaining the circulation? (They are known to be important to momentum transport in the atmospheres of Earth, Mars, and Venus.) Inferences from Atmosphere Structure Experiment data could provide insights into these questions. The vertical lapse rate of temperature dT/dz is a key dynamical parameter which defines the static stability of the atmosphere against convective overturning. Adiabatic lapse rates imply convection; subadiabatic, stable stratification. Lapse rate will be determined as a function of altitude. Discontinuities in the lapse rate will provide information on condensation of condensible species (e.g., ammonia, water vapor, ammonium hydrosulfide) and locate cloud boundaries. Correlation of the measurements of temperature and pressure with the Nephelometer cloud observations will define altitudes and vertical separation of cloud layers, while state properties and lapse rates within the cloud layers will suggest their composition. The magnitude of the vertical winds and their variability, inferred from measurements of descent velocity, dz/dt, could provide direct observations of vigorous convection, or of the presence and amplitude of gravity waves in stable layers. (Gravity waves can cause oscillatory variations in the descent velocity.) The descent velocities are also essential to the analysis of the Probe Doppler Wind Experiment, since they make it possible to resolve the line-of-sight velocity into vertical and horizontal components. Random variations in the instantaneous acceleration of the Probe center of gravity will indicate the presence, scale, and intensity of turbulence in the atmosphere. All of these measurements should provide a valuable 'ground truth' base for use by the remote sensing instruments on the Orbiter, which will then be better able to interpret their global observations. A discussion of the Galileo atmospheric science goals, and a brief description of the Atmosphere Structure Experiment has been given by Hunten et al. (1986). These are ambitious and extensive goals, and to accomplish them, the experiment hardware must perform up to design expectations. To cope with the task of analyzing the data to satisfy the objectives, the following team of experimenters has been formed: A. Seiff, San Jose State University Foundation, PI; D.B. Kirk, University of Oregon; R.E. Young, Ames Research Center, NASA; G. Schubert, University of California, Los Angeles; J. Mihalov, Ames Research Center, NASA; T.C.D. Knight, Martin Marietta Aerospace; R.C. Blanchard, Langley Research Center, NASA; and S.C. Sommer, retired (formerly at Ames Research Center, NASA). Fig. 2a. Fig. 2b. Fig. 2c. Fig. 2. The measurement range and resolution of the Atmosphere Structure Instrument sensors, compared with Orton's nominal model of Jupiter's atmosphere: (a) Temperatures anticipated in the descent mode, from +50 to -150 km. The instrument has been designed to measure from 0 to 500 K, with overflow capacity to 550 K. Estimated accuracy ranges from ~0.1 K at the low end to ~1 K at 500 K (see text). (b) Pressures anticipated in descent mode. Sensor range boundaries and initial resolution on each range are indicated. See text for accuracy discussion. (c) Densities and temperatures expected in the middle and upper atmosphere. Acceleration sensor threshold and ranging boundaries are shown. See text for accuracy discussion. 2.1. MEASUREMENT RANGES, RESOLUTION, AND ACCURACY An indication of measurement data we hope to obtain in descent mode is given in Figures 2(a) and 2(b), which show the instrument ranges compared with profiles of temperatures and pressure from a nominal atmosphere proposed by Orton (1981). (The deep atmosphere is modeled as adiabatic.) Pressure and temperature will be sampled at 2 s intervals in descent. Altitude resolution will be 0.2 to 0.06 km at descent velocities from 100 to 30 m s^-1. As can be seen in the figure, this will give a very dense array of points along the curves (50 to 160 points in 10 km), the number increasing with depth in descent as the Probe descent velocity decreases. Absolute uncertainty of the temperature measurements is expected to be ~1 K at the higher temperatures and < 0.1 K near 100 K (hardly perceptible on the graph). The selected temperature range, 0 to 500 K, provides some margin beyond temperatures expected in descent. The measurement range extends to the 38 bar level in an adiabatic atmosphere. Pressure sensors of three ranges, 0.5, 4, and 28 bars full scale, will be read to 10 bit resolution, giving least count values of 0.5, 4, and 28 mb, respectively. The size of the three dots in Figure 2(b), placed at locations of maximum uncertainty relative to pressure magnitude (low in the three ranges just after range change) indicate the resolution uncertainty. Absolute pressure errors could be as large as 3 counts, three times the dot sizes. Further discussion of accuracy is given below. Figure 2(c) shows the density and temperature profiles from Orton's nominal model of the middle and upper atmosphere, as functions of altitude above the 1 bar level. The exospheric model temperature profile is guided by Voyager UV solar and stellar occultation data, which include a deduced temperature of 1100 +- 200 K at 1450 km altitude (Atreya et al., 1981; Festou et al., 1981). The model temperature variations induce the curvature shown in the density profile. (Conversely, from observed curvature, temperature variations may be deduced.) Notice the comparatively straight section of the density curve where the model atmosphere is nearly isothermal, between 100 and 350 km altitude. The accelerometers to be used to define the atmosphere above 50 km have 4 ranges in the probe axial direction, with a dynamic range from 3 micro g to 400 g. The altitude intervals of use for each range are indicated in Figure 2(c). Altitude resolution is important to defining local curvature in the density profile and short vertical wavelength variations in the profile, such as gravity wave structures, thought likely to be present in the upper atmosphere. Altitude resolution is determined by sampling frequency. This was limited by the available Probe memory capacity, since the entry data are stored for transmission in descent. Samples will be taken at fixed time intervals of 5/16 s. alternating between the primary z1 sensor and the redundant or secondary sensor z2. This leads to altitude resolution of ~2.0 km throughout ranges 1, 2, and 3, in which, because of the low ambient density, velocity and path angle do not change much from values at entry. At 100 km altitude (near peak deceleration), velocity has decreased, and altitude resolution is improved to ~1.0 km. Altitude resolution continues to improve to a value of ~0.15 km just prior to parachute deployment near 50 km altitude. The resolution above 200 km will permit waves of vertical wavelength ~10 km to be detected and defined. Wavelengths longer than 1 km vertically will be defined at 50 km altitude. The measurement resolution of the axial accelerometers on the four ranges is nominally 3 micro g, 0.1 mg, 3 mg, and 0.1 g, respectively. We take the range 1 threshold to be 15 micro g (5 counts), which corresponds to 1.6 x 10^-11 kgt m^-3 atmospheric density. In the model atmosphere, Figure 2(c), this places the threshold at 750 km altitude where density is defined within the +- 10 percent resolution (indicated by the bar height at the threshold point). Measurement resolution improves to +- 1 percent at 615 km altitude, and, at full scale on ranges 1, 2, and 4, to 0.025 percent of reading (0.05 percent on range 3). After upranging, on each new range, initial resolution is 0.8 percent. Altitudes at which ranging will occur are indicated in Figure 2(c). Basic sensor accuracy is ~0.01 percent, so resolution, not accuracy, will always be the limitation. Atmospheric temperature uncertainties from such a data set were shown for Pioneer Venus data (with less capable accelerometers and less frequent sampling relative to acceleration pulse duration) to be typically a few K, with maximum uncertainty ~10 K immediately after range change. However, this does not include the uncertainty in mean molecular weight of Jupiter's middle and upper atmosphere. (The mean molecular weight mu is required in the equation of state to derive temperature from density and pressure.) Errors in the molecular weight lead directly to the same fractional error in temperature. There is no present basis for estimating this uncertainty quantitatively, but we expect it will be < 5 percent and possibly as small as 1 percent for the middle atmosphere, and < 10 percent for the upper atmosphere, after the complete set of investigations to be performed by the Galileo Probe (including a neutral mass spectrometer and a helium abundance detector) and the Orbiter (including a UV spectrometer) has been analyzed. 3. Sensors Sensors selected to make the above measurements are described here, along with expected accuracies and principal sources of error. 3.1. TEMPERATURE SENSORS The temperature sensors for Galileo were modeled after those used on the four probes of the Pioneer Venus Mission. They were designed for fast response (i.e., good thermal coupling to the atmosphere and small thermal inertia), and to be insensitive to support conduction, thermal radiation, and self heating errors (Seiff, 1976; Seiff et al., 1980b). The sensors are deployed outside the probe boundary layer in a region where local flow velocity around the probe is high, to avoid thermal contamination and promote rapid heat transfer. Figure 3 shows their location, as well as the locations of the pressure sampling inlet, the accelerometers (near the Probe center of gravity) and of the electronics and pressure sensors within the Probe. Figure 4 is a schematic of the temperature sensor, which is a dual element platinum resistance thermometer. Primary and secondary sensors, designated T1 and T2, are mounted on a single head. The primary sensor has very short response time and minimizes thermal errors. The secondary sensor has acceptable response and error rejection, and is better protected, less susceptible to damage in handling and in flight. Both were demonstrated by tests to survive the specified engineering environments. (Both survived the launch, entry, and descent environments on the Pioneer Venus Mission.) Both should measure temperatures accurately under conditions of descent in Jupiter's atmosphere. The primary (T1) sensor is a fine platinum wire directly exposed to the atmosphere. It is wound around an open frame of platinum-rhodium alloy, with atmospheric gases flowing through the frame (Figure 4). The wire is 0.1 mm in diameter by approximately 1.2 m long. It is insulated from the frame by a thin layer of glass, and is fixed in position by a second layer of glass along the outer posts. The secondary sensor (T2) is a 6 cm length of 0.025 mm diameter wire configured as a 1 cm long raster over a thin glass film on the leading side of the outermost tube in the sensor frame (see enlarged inset in Figure 4). It is covered with a thin, protective coating of glass. To prevent resistance changes due to thermal strain, the glass used in the coatings was selected to match the thermal expansion coefficients of the frame and sensing elements. Any residual strain effects were included in the calibrations over the full use range of the sensors. The sensor support stem is a thin-walled tube of low-conductivity stainless steel, which limits conduction between the sensor and the Probe shelf to which it is mounted. Fig. 3. Installation of the ASI on the Galileo Probe descent configuration. The temperature and pressure sensors sample flow outside the descent module boundary layer. The Probe center of gravity lies within the accelerometer package. Pioneer Venus experience* led us to coat all external surfaces of the sensor with a 25 micro meter thick insulating polymer film (vapor deposited Parylene) in order to prevent shorting of the sensor by cloud materials. The polymer is, in turn, covered by a 1000 A film of gold, to ground the external surface to the Probe and prevent potential differences which could * Partial shorting of the T1 sensors occurred during passage through the clouds of Venus, due to accumulation of a conductive film of sulfuric acid on the wire surfaces. The shorting cleared below the clouds, as the film was blown off by atmospheric flow (Seiff et al., 1980a). lead to static discharges. The low emissivity of the cold coating also limits radiative exchange with the environment. A second modification to the Pioneer Venus design was the provision of an electrostatic discharge shield around the sensor. The shield, shown in Figure 4(b), is about 4.5 cm in diameter and is open at front and rear, so that the atmosphere flows through freely. Wires cross the openings at about 1 cm intervals, to enclose the sensor in a space at Probe potential. If, during descent, an electric charge builds up on the Probe, or if electrostatic fields in the atmosphere cause discharges to take place, the cage is intended to prevent currents from discharging to the sensor. Tests of the shielded sensor were made at reduced pressures in a bell jar. Ten thousand volt discharges were created between the sensor assembly and a plate nearby. The discharges always took place to the shield, and sensor resistance measurements Fig. 4a. Fig. 4b. Fig. 4. The temperature sensor. (a) Sketches showing the sensor configuration and mounting. The sensor is mounted to the Probe instrument shelf. Its stem passes through the Probe wall and boundary layer to deploy the sensor into external atmospheric flow. (b) View of the sensor enclosed in the electrostatic discharge shield. taken simultaneously were found to be unaffected by the discharge. The shield could be important if the Probe encounters strong electrostatic fields or lightning discharges in the atmosphere of Jupiter. The sensors were fabricated and calibrated by Rosemount, Inc. They were calibrated by immersion in baths of liquid nitrogen (77 K), liquid oxygen (89 K), ice water (273 K), boiling water (373 K), and against a secondary standard at 477 K. The sensors were enclosed in plastic bags and were fully immersed in the calibration bath and left until resistances were stable. The resistances at known temperatures were then used to calculate constants in the familiar Callendar-Van Dusen equation (see, e.g., Riddle et al., 1973) for purposes of interpolation between calibration points above 273 K. (The resistance function R(T) of platinum is the basis of the International Practical Temperature Scale.) Below 273 K, data were referenced to the International Practical Temperature Scale (68), and correction curves for deviations from that scale were derived (Rosemount Report 27410, 1975). Sensor calibrations were checked several times at Ames Research Center over a period of years. The initial Ames calibration was within a few hundredths K of the Rosemount calibrations in liquid nitrogen. The repeatability of the Ames calibrations is shown in Figure 5 for two temperatures, 77.4 K (liquid nitrogen) and 273.15 K (ice bath). At 77K, the T1 calibration data group very closely around two lines differing by 0.15 K, which have the theoretical slope dR/dT for platinum. The standard deviation of the points from their associated lines is 0.017 K. The left-hand line fits calibrations taken before November 1983, at which time the sensors were coated with parylene and Fig. 5a. Fig. 5b. Fig. 5. Temperature sensor calibration data taken over the period from 1983 to 1988. (a) In liquid nitrogen baths. (b) In water ice baths, gold. The right-hand line fits two calibrations taken subsequently. It appears that the T1 calibrations at 77 K were measurably affected (by 0.15 K) by application of the coatings, and the 1984, 85 calibrations will be used in reducing flight data. In the lower figure, the full set of T2 calibration data in liquid nitrogen shows a standard deviation of 0.13 K about the solid line. This sensor required a repair in March 1983, which increased its resistance at 0 degrees C by 0.12 Ohms (1 percent). To take the effect of this change out of Figure 5(a), the resistance at 77 K has been ratioed to its value at 273.15 K. The data fall into 3 groups, one in early 1983, one after the repair (late 1983), and the third after coating application in November 1983. The dashed line of theoretical slope through the two data points taken in 1984 and 1985 (about 0.13 K to the right of the solid line) is considered the calibration of record. Calibrations of sensor resistance in an ice bath over the period from 1983 to 1988 are shown in Figure 5(b) to an enlarged vertical scale. Mean values are shown by horizontal lines. Standard deviations about the mean are 0.069 K for T1 and 0.13 K for T2 over the 5-year period. A rising trend in resistance with time after May, 1984 is suggested by 3 data points. We can see no physical cause for this, and do not believe is real. However, we shall be alert to this possible trend in reducing the data at Jupiter. The bars are 0.1 K and 0.25 K high. To functionally test the sensor over the full range to 500 K and to compare T1 and T2 readings against one another more generally and against some calibrated thermocouples, a thermally insulated, closed circuit flow channel (a table-top wind tunnel) was built, in which gas (either air or helium) was circulated at atmospheric pressure. Electrical heaters on the backside of the circuit warmed the gas at a rate to match that expected in descent on Jupiter. The sensor projected through the channel wall into the flow about as far as it extends outside the descent module into the atmosphere, when mounted on the probe. The channel was made of ordinary 6-inch diameter, sheet metal duct pipe, insulated to reduce heat exchange with the ambient atmosphere. The two long legs of the channel were about 1.5 m long; the short legs, about 0.75 m long. A fan, mounted internally, circulated the flow at a velocity of 6 m s^-1. It was found that there was considerable thermal inhomogeneity in this channel during the heating cycle, and the best data were taken when the environment was cooling from 500 K back to room temperature. Under these conditions, the following results are obtained: (1) T1 and T2 agreed typically within about 0.2 K at 450 K, their difference decreasing with temperature. (This was without correction for the lag of the T2 sensor, which was greater in the test than it will be on Jupiter because of differences in gas density and flow velocity.) (2) The temperatures agreed with those given by the closest nearby thermocouples within 0.1 K typically. (3) Tests made with and without the electrostatic discharge shield showed that readings were not affected significantly by its presence. Calibrations run end-to-end with the ASI sensing electronics, also over a period of years, were consistent with analog, sensor level calibrations within 1 count. The above data suggest an accuracy of calibration of about 0.1 K T1 and 0.2 K T2 at 500 K, the absolute error tending to increase with temperature. However, in order not to promote extreme expectations, we presently suggest an uncertainty of 1 K at 500 K, and 0.1 K at 100 K when all other error sources are present. We will refine these uncertainties as warranted by the data returned. Relative accuracy over short time spans should be limited only by the resolution, 0.013 K. Response time varies with ambient density and descent velocity, from 16 ms at deployment to 5 ms at 16 bars for the primary sensor, and from 300 ms to 80 ms for the secondary. The sensors will track the rise of atmospheric temperature during descent, T1 to within 0.002 K, and T2 within 0.038 K. A more severe requirement on response time is imposed by the goal of measuring turbulent temperature fluctuations. With the response achieved, turbulent fluctuations of 100 m scale and 1 K amplitude, for example, will be followed by the primary sensor within about 0.01 K at a descent velocity of 100 m s^-1. The sensors will be sampled alternatively, at 2-s intervals. If either sensor fails, failure logic in the microprocessor provides that data from only the surviving sensor is to be taken at 2-s intervals. The failure logic simply looks for an open or shorted sensor, as indicated by overscale or zero readings which persist for one major frame (a 64-s period). 3.2. PRESSURE SENSORS The pressure sensors in the Galileo Atmosphere Structure Instrument measure pressure from the deflection of a thin, stainless steel diaphragm, 1.5 cm in diameter, with a sealed vacuum reference on its backside. (The sensors were provided by the Tavis Corporation, and are commercially available.) A sensor of this type was used in the parachute phase of the Atmosphere Structure Experiment on each Viking Lander, and, after landing, was handed off to the Meteorology Team to measure daily and seasonal pressure variations on Mars. It did so successfully over a period of years. The diaphragm deflection is measured by changes in reluctance of magnetic circuits, which include the air gaps between the diaphragm and sensing plates on either side of the diaphragm. The sensor external dimensions are 2.5 cm diameter by 10 cm long. Each sensor weighs 129 g and requires 150 mW of power. There are three sensors in the instrument, with full scale ranges of 500 mb, 4 bars, and 28 bars. (This geometric range scaling, by factors of 8 and 7, maintains roughly constant percentage reading accuracy over the descent.) The maximum pressure of 28 bars was selected to provide data in the event that the Probe survives to greater than the design depth on Jupiter. This sensor can also run overrange by about 100 percent. The three sensors can be seen mounted to the electronics housing and connected to the inlet manifold tubing in Figure 1(a). On the Probe, the manifold leads to a pitot tube inlet located along the side of the descent module, outside the Probe boundary layer, Figure 3. In Figure 1(b), the sensors and inlet are enclosed in a thermal insulation blanket (see below). The signal level is 0 to 5 V full scale. With 10 bit A/D conversion, the resolution is 1/1024 of full scale on each range. Resolution uncertainty of a single reading thus varies from 0.5 percent of reading at first use to 0.1 percent at full scale. Sampling is simultaneous with temperature sampling, at intervals of 2 s. At p < 500 mb, ranges 1 and 2 are sampled alternately. At 0.5 < p < 4 bars, ranges 2 and 3 alternate. This sampling scheme provides redundancy and measurement accuracy checks, within the resolution of the higher range sensor. Above 4 bars, where only the 28 bar sensor is on range, it is sampled every 2 s. Failure detection logic was provided for the two lower ranges. The sensor is considered to have failed if it gives zero output for one major frame (64 s), or if the reading is constant over one major frame. Failed sensors are checked continuously and reinstated if the above conditions disappear. Another failure protection implemented is that the lowest range sensor will not be read after 6 major frames into the descent, and the middle range will not be read after 28 major frames. These sensors should be well over range after these periods of descent, with deployment at 100 mb. Accuracy and stability of the sensors were evaluated by repeated calibrations over a period of years while Galileo was awaiting launch. Typically, scale factors at a given temperature were found to be stable within 0.1 percent over these years. End-to-end calibrations, taken through flight electronics, were consistent with sensor level calibrations, within the digital resolution. Sensor scale factors vary with temperature by 0.1 percent to 0.5 percent over the temperature range from -20 degrees C to +50 degrees C. The temperature dependence was repeatable within narrow limits over a period of years, and is part of the calibration data set. The sensor offsets vary with temperature to the extent of 20 to 50 mV (0.2 percent to 0.5 percent of the 5 V full scale signal) over the same temperature range. These variations are repeatable within 10 mV. The signal at a given pressure is also sensitive to the rate of change of probe internal temperatures. It was to minimize this effect that the sensors were thermally insulated. The remaining magnitude of the effect is ~25 mV maximum in the nominal descent thermal environment, in which probe internal temperature changes at the rate ~1 degree C min^-1. Corrections to the data will be made for these thermal offsets. To define the corrections, the Probe interior temperature history descent was simulated, and offsets induced were recorded. Tests were performed in the nominal and in a worst case environment with dT/dt up to 2.9 degrees C min^-1. The offsets were found to be correlated with the temperature difference between sensor internal thermometers and one mounted on the manifold coupling of the range 3 sensor. These four temperatures will be read as engineering data in flight. The correction is expected to be accurate within 5 mV (1 count or 0.1 percent of full scale signal) in the nominal descent environment. Sensor offsets were typically set at 100 to 200 mV. They differ by about 10 mV from turn-on to turn-on. While this is undesirable, it is not expected to affect accuracy seriously, since offsets will be read just prior to entry into the atmosphere, and should be predictable through the ensuing hour of descent. At worst, it is estimated that offset uncertainty could introduce 1 or 2 counts of uncertainty (out of 1024, full scale) in the data. The effect on sensor output of Jupiter's strong intrinsic magnetic field, which is the order of 10 x that of Earth, was investigated and found to be significant. Pioneer 10 and 11 data (Smith et al., 1976) indicate a magnetic field strength within Jupiter's equatorial region of approximately 4 G. The sensors were tested in magnetic fields up to 12 G. Sensor scale factors were unaffected by the applied field, but offsets were found to be sensitive to both axial and transverse fields by as much as 50 to 200 mV. This was not originally anticipated, but it is not surprising, considering that the diaphragm deflections are sensed magnetically. Offset shifts induced in these tests remained after the test field was turned off. To preserve the planned accuracy, the sensors were magnetically shielded in 'mu-metal' housings 0.5 mm thick, which extend over the sensing heads. (Shielding material was supplied and housings were fabricated by Advance Magnetics, Inc.) The shields reduce the field within by a factor of 20. With shields in place, sensor outputs were constant within 1 count (5 mV) in both transverse and axial 12 G fields, while the sensors were rotated through 360 degrees. 3.3. ACCELERATION SENSORS The acceleration sensors on the Galileo Probe will play an important role by defining the structure of the upper and middle atmosphere (pressure levels 10^-7 to 100 mb). Decelerations measured in the probe axis of symmetry (z-axis) provide the basic data, while the lateral (x and y) axis sensors define the Probe angle of attack history during entry and thus the magnitude and direction of the resultant aerodynamic force. In the lower atmosphere, accelerometers will establish vertical wind magnitude, and characterize turbulent motions in the atmosphere, in both axial and lateral directions. To respond to these objectives, the sensors must have a broad dynamic range. In descent, measured axial accelerations will be close to Jupiter's gravitational acceleration, nominally 2.37 g (g is Earth's gravitational acceleration, 9.806 m s^-2) . During entry, the nominal deceleration peak is 260 g, with an upper limit near 400 g. The minimum detectable acceleration determines the experiment threshold, and for this purpose, a sensitivity of 3 micro g count was selected, based on knowledge of capabilities of existing sensors. (Tests of the Galileo sensors indicate they are indeed capable of resolving < 3 micro g on the most sensitive range.) Thus, the dynamic range extends from 3 micro g to 409 g. The accelerometers were designed and manufactured by the Bell Aerospace Division of Textron. They were derived from Model XI guidance accelerometers, modified to meet specific experiment requirements. (Bell Model IX sensors were used in the Viking instrument.) A key modification was extension of maximum measuring range. External and sectional views of one sensor are shown in Figure 6. A photograph of the four-sensor flight unit is included in Figure 1(b). Fig. 6. Schematic and external view of one acceleration sensor. The operating principle of the sensors may be understood from the sectional drawing in Figure 6. When the sensor experiences acceleration, its electronics detect movement of the flexure supported test mass from its null position (by capacitive sensing) and supply a servo-controlled current to the coil in the test mass to restore it to the null position. The coil field interacts with the intense field of the permanent magnets to provide necessary restoring force. The current in the coil is then the measure of the acceleration. The servo system constrains the test mass to its null position within 1 microrad g^-1. Thus, maximum deflection at 410 g is 0.4 mrad, corresponding to test mass linear displacement < 10 microns. This leads to exceptionally linear response to input accelerations over the full measurement range. The sensor analog electronics generate a signal of 0 to 5 V, or, where both positive and negative accelerations must be sensed (x- and y-axes, and z-axes on Range 3), +-2.4 V. Individual sensors fit within a 2.3 cm diameter x 3.4 cm long envelope (Figure 6) and weigh 48 g. The four sensors are rigidly mounted on a bracket, with the z-axis sensors on the probe axis (Figure 2), and the z1 sensor very near the probe center of gravity during entry. The displacement of the lateral axis sensors from the Probe axis of symmetry is normal to the directions of their sensitive axes. Thus, the only rotational acceleration to which they respond is spin acceleration. The weight of the assembly is 540 g, much of the weight being in the bracket. Power required varies from 1.25 W at 0 g to 6.5 W at 410 g. A single cable brings conditioned power and accelerometer range change commands from the ASI electronics, and conducts analog signals from the 4 sensors to the ASI electronics (Figure 1(b)). The two z-axis sensors operate on 4 ranges with full scale levels of 12 mg, 0.4 g, +-6.4 g, and 410 g. The x- and y-axis sensors have 3 ranges with full scale levels of +-12.5 mg, +-8 g, and +-12.8 g. The digital resolution is 12 bits in the z-axis, 8 bits in the two lateral axes. Thus, in the z-axis, 1 count ranges in value from 3 micro g to 0.1 g. Command capability was provided to choose how the two z-axis accelerometers will be used during Jupiter entry and descent, depending on how well each functions during interplanetary cruise. Either can be used alone, or both can be used with samples interdigitated. Either can also be selected as the 'primary sensor', the source of the z-axial turbulence data. The data taken during the high-speed atmosphere entry have to be stored because of radio communication blackout induced by the plasma sheath around the Probe. The data are stored in the Probe memory and will be read out twice from designated words in the Probe data format during descent below the 0.1 bar level. The memory capacity provided was 12 kbits. This allocation, over the estimated maximum entry period of 240 s, allows a data collection rate of 50 bps, which had to be carefully budgeted. Thirty-six bps were allocated to axial decelerations, the basic data which define the density profile. A 12-bit sample is taken from each z-axis sensor every 5/8 s. Samples from the two sensors are staggered, so that an axial acceleration is read every 5/16 s. To conserve data, the resultant acceleration normal to the z-axis, termed a_n is computed from a_x and a_y in the ASI microprocessor, and one sample/second is transmitted at 8-bit resolution. This is to be used to define the probe resultant angle of attack oscillations from observed oscillations in the lateral acceleration. The remaining 6 bps are used for engineering data, sensor ranges and temperatures. In Descent Mode, the total experiment data rate is 18 bps. Of this, 6 bps are allocated to temperature, 6 bps to pressure, 4.5 bps to acceleration data, and 1.5 bps to engineering data. Within the acceleration data, 1.5 bps are used to define the total velocity change in the z-axial direction, as derived from each of the z1 and z2 sensors over staggered 16 sec intervals; 1.5 bps, to transmitting the mean, maximum, and minimum value of a_n in 16 s intervals; and 1.5 bps to transmit statistics on the axial and lateral turbulence levels every 96 s. The total velocity change in the z-direction is integrated by the microprocessor which samples the accelerations 32 times s^-1. Thus, it is evident that the use of data is very economical, and minimal. It is also evident that the data collection pattern is moderately complex, but is readily handled by programming the experiment control microprocessor. A more complete exposition of the sampling sequences and their implementation is given in the following section. Turbulence data are compressed by counting excursions from the mean acceleration which exceed pre-specified levels, both in the z-axis and normal to the z-axis. The pre-specified deviation levels in a_z are +-0.025 g_J, +-0.05 g_J, +-0.10 g_J, and +-0.20 g_J (where g_J is the acceleration of gravity on Jupiter, taken as 2.37 g_E). The pre-specified levels in a_n are 0.0019, 0.0056, 0.0113, and 0.0226 g_J. The number of excursions of the acceleration above each of these levels is counted in a 96-s sample period, and the number of counts is transmitted in 12, 12-bit words. (Positive and negative excursions in z are independently counted, while only the absolute magnitude of the excursions is read on n.) The interval mean acceleration in z is taken to be the average value in the 16-s period preceding the 96-s sample. In the two transverse axes (and hence in a_n), the Probe is assumed to be symmetric and the acceleration is assumed to be zero in the absence of turbulence. The acceleration sensors were calibrated against the Earth's gravitational acceleration by rotation through 360 degrees under temperature controlled conditions at Bell Aerospace in Buffalo, New York. The sensor level calibrations were done 3 times, before and after environmental testing (in December 1983 and February 1984), and on return of the instrument to Bell for a minor repair in mid-1985. The scale factors were defined to 5 significant figures, and they repeat within 0.3 percent on Range 1 (most sensitive) and 0.01 percent on Range 3 (6.5 g full scale). The sensors are temperature compensated, but residual sensitivity was determined by calibration at four temperatures. End-to-end calibration checks have been repeated many times in the Earth's gravity field, the last being in July 1988, and they show no changes with time in the scale factors, within the digital resolution. The bias values have long term stabilities of about 150 micro g's on all ranges, and the values in effect at the time of Jupiter encounter will be read at instrument turn-on prior to entry. It was considered unsatisfactory to calibrate the scale factor on the 410 g range with gravitational inputs limited to +- 1 g. As an alternative, precisely known stimulus currents were applied to the coil in the test mass to simulate input accelerations up to 410 g. The sensor current scale factors (ma/g) determined to 5 significant figure accuracy on Range 3 at 1 g input, were used to calculate acceleration values equivalent to the stimulus currents. It was found that the Range 4 scale factors so determined (g count^-1) checked those obtained from +- 1 g calibrations within about 0.16 percent. The sensor calibration on Range 4 was also checked by centrifuge tests with the complete ASI operating at inputs up to 210 g. There was agreement within 0.1 percent between sensor acceleration data and input accelerations determined from centrifuge rotation rate. (This is the limit of accuracy to which input accelerations could be determined.) This test was a simulation of the full use cycle of the instrument during entry into Jupiter's atmosphere, and showed satisfactory range switching operation, and also tested the agreement between the z1 and z2 sensors. 4. Instrument Electronics Design A critical factor in the electronics design was the fact that the Probe passes through the Jovian radiation belts prior to entering the atmosphere. As a consequence, the instrument has to operate within specification after exposure to a significant radiation dose. Parts selection was severely constrained, radiation testing was performed on samples from every lot of parts used, and the potential effects of part parameter variation with radiation dose had to be considered along with other parametric variations in designing for stability of electronic circuit operation. In general, parts were selected that can be exposed to doses of 10^6 rad (Si) without unacceptable performance degradation. Use of a microprocessor for control of this instrument allowed significant flexibility in instrument operation, data processing and output data formatting. In addition, it permitted several failure sensing algorithms to be implemented, to sense gross sensor failures and substitute data from healthy sensors. The flexibility of operation made it necessary for the ground support equipment to be rather sophisticated in order to be able to exercise all possible operational branching paths within the instrument software. The basic instrument electronics system design is shown in Figure 7. This design, completed during 1979, is centered around an RCA CMOS microprocessor system comprising the CDP 1802 central processing unit (CPU), 8192 8-bit words of read-only memory (ROM), and 256 8-bit words of read/write random access memory (RAM). The system is clocked at about 2.2 MHz, a low frequency used to retain an acceptable operating margin after radiation exposure degradation. The ROM contains the instrument's operating and control software and the RAM provides for temporary storage during data accumulation and manipulation and also provides the data output buffer. Control signals from the Probe (comprising the operating mode commands, axial accelerometer selection commands, data frame markers, data transfer clocks, and the Probe clock) are used to control the operation of the instrument by generating system interrupts. Each system interrupt causes the microprocessor to suspend the sequence in process, interrogate the interrupt registers, execute the response, and return to the sequence in process, if appropriate. In the case of an operating mode command, the microprocessor branches to the correct mode, reinitializes, and proceeds in the new operating mode at the next Probe minor frame interrupt. For other interrupts, the microprocessor completes the interrupt action and then returns to where it was prior to the interrupt. The instrument is synchronized and slaved to the Probe timing as a result of this interrupt control. The control signals from the Probe and the functions performed are shown in Table I. The design incorporates two 12-bit analog-to-digital (A/D) converters of the dual slope integrating type, each with an auto-zeroing loop. The A/D converter precision reference voltage is derived from a highly stable zener diode having internal temperature compensation. This same reference voltage is used to derive the precision stimulus Fig. 7. Atmosphere Structure Instrument electronics block diagram. Table 1. Probe/ASI control signals and functions. currents and voltages for the sensors, including the constant current drive through the platinum resistance temperature sensors, the calibration currents supplied to the accelerometers, and the calibration voltages supplied to the pressure sensors. This approach results in an in-flight instrument calibration that is, to a first approximation, insensitive to the absolute stability of the reference source. The accelerometers and pressure sensors provide 0 to +5 V outputs to the central electronics. For the platinum resistance temperature sensors, however, the central electronics provide the stimulus currents and signal amplification. The simplified temperature measurement implementation is shown in Figure 8. A four wire configuration is used for each sensor, to eliminate measurement errors caused by voltage drop in the lead wires between the central electronics and the sensors. One pair of wires carries the stimulus current while the voltage measurement across each sensor is made through the second pair of wires, which look into the high impedance of the signal amplifier. The Fig. 8. Temperature measurement implementation. configuration also provides a zero-offset determination and a scale factor calibration, by means of a stable, 12 Ohm calibration resistor carried in the electronics. The analog outputs from the sensors, conditioned as necessary, are multiplexed to the A/D converters under control of the sequence programmed into the microprocessor. Data sampling is slaved to the 2048 Hz probe clock, which is divided down to generate a data sampling interrupt at the rate of 128 samples per second. The microprocessor responds to the data sampling interrupt by setting appropriate system flags that are used later in processing the data. The data sampling interrupt itself initiates the 'start A/D conversion' in hardware, to avoid the possibility of small but variable delays in initiating it through the software interrupt process. The basic sensor sampling sequence is a four data sampling interrupt cycle, as shown in Table II. Each of the accelerometer outputs is sampled 32 times per second, with the TABLE 2. BASIC SAMPLING SEQUENCE CYCLE data samples for the pressure, temperature, and housekeeping sensors interspersed as necessary. The accelerometer sampling rate is the same in all operating modes and all of the data processing, including integration of acceleration to provide velocity change data, the computation of the resultant of the two lateral accelerometer samples of every measurement, and rescaling of data at accelerometer range changes, is performed digitally within the microprocessor system. The logic for determining when to make accelerometer range changes, and sending the range change commands is also implemented within the microprocessor. Failure sensing techniques are applied to the temperature and pressure sensors to enable substitution of data in the event of a gross sensor failure, as discussed earlier. The three instrument operating modes, calibration, entry, and descent may be functionally described as follows: The calibration mode is used at a time when the probe is still in free fall after passing through the radiation belts. It provides zero-offset and some scale factor calibrations of the sensors and will also calibrate any electronics changes. The entry mode sampling strategy emphasizes deceleration measurements, appropriate for the high speed ballistic entry portion of the mission (see above). In the descent mode, atmospheric temperature and pressure become the primary measurements. Because of the limited data rate of the Probe communications link (128 bits per second), the data transfer and transmission rates allocated to the ASI are low, 18 bits per second in the descent mode (about 1/7 of the Probe link capacity) and 50 bits per second in the entry mode, when data go into Probe memory for readout in the hour of descent. These rates critically affect the selection of the data parameters to be output, shown in Tables III and IV. The data output of 18 bits per second in the calibration mode is not critical because the quantity of data to be collected is limited, and there is no severe constraint on the collection interval, since these data are taken at a time before science data gathering begins, and are stored in the Probe memory. The parameters measured during the calibration sequence are shown in Table V. The data output for each calibrate mode parameter is the average of 16 separate measurements of that parameter. Data are taken relatively quickly, averaged, and stored in the instrument buffer memory while waiting for transfer to the probe. The zero-offset data for each sensor are also retained in instrument memory and multiplexed, as 16-bit accumulations, into the ASI descent mode data stream. Some of the calibration data are required during on-board data processing, in both the entry and descent modes. Default values for these parameters are programmed into the microprocessor system to guard against a Probe failure in which a calibration sequence is not run. The instrument software was written directly in the microprocessor machine code instruction set. This was necessary in order to optimize the processing and achieve the necessary processing throughout at the modest 2.2 MHz system clock frequency. Packaging of the electronics used conventional techniques for this type of instrument. The parts were installed on custom designed multilayer (up to 9 layers) printed circuit boards. The packaging was designed and tested to survive a 400 g deceleration along the flight path axis as well as normal launch environments. The other somewhat unusual requirement was to survive the high pressure of the Jovian atmosphere during descent. In order to avoid qualification problems with the large flatpacks being used, the approach taken was to seal the housing containing the electronics, initially at 1 atm of air pressure. The housing was subsequently tested over a 1-h period with external pressure gradually increasing to 20 bar to simulate descent on Jupiter. This test, performed three times over the years, showed small, but acceptable leakage. Based on these tests, the housing is expected to be evacuated in over 6 years in space, and to remain at pressures below about 0.35 atm during descent in Jupiter's atmosphere. There are no high voltages used in the electronics, and hence no danger of arcing at low internal pressures. The instrument was tested at vacuum to verify this. TABLE 3. ENTRY MODE DATA OUTPUTS TABLE 4. DESCENT MODE DATA OUTPUTS TABLE 5. CALIBRATION MEASUREMENT SEQUENCE 5. Concluding Remarks In the design of the Atmosphere Structure Instrument for the Galileo Probe, the goal has been to obtain accurate and detailed information on the structure of the atmosphere from the exosphere to the maximum attainable depth in the lower atmosphere, 10^-7 mbar to 16 bar, nominally. A corollary goal is to define properties of the atmosphere important to atmospheric dynamics. Sensor selection was based on prior experience with conceptually similar instruments in the atmospheres of Mars and Venus, with modifications to avoid problems experienced in earlier missions and those anticipated in Jupiter's atmosphere. The electronics are microprocessor controlled and have added capabilities to those used previously. They permit versatile sampling sequences and make maximum use of the limited available data rate by means of some on-board data processing. If the instrument performs up to design expectations, it should return valuable information about structure and dynamics of the largest planet in the Solar System. References Atreya, S.K., Donahue, T.M., and Festou, M.C.: 1981, 'Jupiter, Structure and Composition of the Upper Atmosphere', Astrophys. J. 247, L43. Festou, M.C., Atreya, S.K., Donahue, T.M., Sandel, B.R., Shemansky, D.E., and Broadfoot, A.L.: 1981, 'Composition and Thermal Profiles of the Jovian Upper Atmosphere Determined by the Voyager Ultraviolet Stellar Occultation Experiment', J. Geophys. Res. 86, 5415. Hunten, D.M., Colin, L., and Hansen, J.E.: 1986, 'Atmospheric Science on the Galileo Mission', Space Sci. Rev. 44, 191. Ingersoll, A.: 1981, 'Jupiter and Saturn', Sci. Amer. 245, 90. Ingersoll, A.: 1990, 'Atmospheric Dynamics of the Outer Planets', Science 248, 308. Orton, G.S.: 1981, Atmospheric Structure in the Equatorial Region of Jupiter, Galileo Project Document 1625-125. Riddle, J.L., Furukawa, G.T., and Plumb, H.H.: 1973, Platinum Resistance Thermometry, NBS Monograph 126, U.S. Dept. of Commerce. Seiff, A.: 1976, 'The Viking Atmosphere Structure Experiment - Techniques, Instruments, and Expected Accuracies', Space Sci. Instr. 2, 381. Seiff, A. and Kirk, D.B.: 1977, 'Structure of the Atmosphere of Mars in Summer at Mid-Latitudes'. J. Geophys. Res. 82, 4364. Seiff, A., Reese, D.E., Sommer, S.C., Kirk, D.B., Whiting, E.E., and Niemann, H.B.: 1973,'PAET, An Entry Probe Experiment in the Earth's Atmosphere', Icarus 18, 525. Seiff, A., Kirk, D.B.. Young, R.E., Blanchard, R.C., Findlay, J. T., Kelly, G.M., and Sommer, S. C.: 1980a, 'Measurements of Thermal Structure and Thermal Contrasts in the Atmosphere of Venus and Related Dynamical Observations: Results from the Four Pioneer Venus Probes', J. Geophys. Res. 85, 7903. Seiff, A., Juergens, D.W., and Lepetich, J.E.: 1980b, 'Atmosphere Structure Instruments on the Four Pioneer Venus Entry Probes', IEEE Trans. Geosci. Rem. Sens. GE-18, 105. Smith, E.J., Davis, L., Jr., and Jones, D.E.: 1976, in T. Gehrels (ed.), 'Jupiter's Magnetic Field and Magnetosphere', Jupiter, University of Arizona Press, Tucson, p. 788.