STRV 1B: EXPERIMENTS

Cryo-cooler Vibration Suppression Experiment

       Cryocoolers are typically used to cool IR or CCD sensors to lower the thermal noise threshold. Stirling coolers have a mechanical compressor. so there is a significant vibration problem with optical instruments. This BMDO experiment uses a novel vibration suppression technique comprising piezo materials to make the tip of the coldfinger stand stilk thus eliminating motion of the sensor mounted on the coldfinger. The objective of the experiment is to demonstrate the design and qualification of low-power flight piezo drivers and control systems, and the effectiveness of this mode of vibration suppression when used in a zero-g, mass-limited system.

       Two actuator methods were used in the experiment : piezo translators and applique ceramics. Commercially available low voltage piezo translators displace the entire cryocooler to cancel the motion of the tip of the cold finger, and three eddy current transducers detect tip motion. In addition, high voltage (700v) applique ceramics are glued to the coldfinger and stretch it to cancel the tip motion. Both analogue and digital control systems were used for comparison of techniques and effectiveness. The cryocooler selected for the experiment is the Texas Instruments 0.2W "tactical" cryocooler. The major components of the experiment are shown in Fig. 16 (above left).

       Prior to launch, problems in implementing the design of the electronics for the high voltage ceramic applique technique resulted in elimination of this approach in flight. It also became clear that the digital control approach was much more effective than the analogue approach. Primarily this was a result of the available resources only allowing for the design of the analogue controller to handle a single harmonic.

       The translators with digital control reduce the vibration associated with the first eight harmonics by a factor of approximately 75. Using analogue control, vibration levels at the first harmonic frequency are reduced by a factor of 10. A fundamental experiment limit is the 10 nm resolution of the eddy current transducers. The results obtained with digital control are at this-limit. Further reduction in tip motion would require more sensitive means of measuring the motion.

       The analogue control circuit is limited by stability considerations to operate only at the fundamental frequency. Although higher gain could be used, the amplitude of motion at the fundamental frequency has been reduced well below neighbouring harmonics. and thus luther reductions in motion at the fundamental frequency would do little to reduce total motion. During ground tests, large commercial high voltage drivers were used with the breadboard electronics to power the flight ceramic applique. Using the analogue control circuitry, the fundamental was reduced by the factor of ten mentioned above. Unfortunately, it was not possible to complete development of the flight electronics package, and thus digital control was the only modality possible after launch.

       After allowing the spacecraft and cryocooler to fully outgas, the cryo-cooler experiment was first operated on 12th July 1994. To date, the cryocooler has been turned on about 1,000 times. and total operating time is about 35 hours. On six occasions the cooler has been operated at temperatures low enough to verfy that it is cooling properly, with tip temperatures <120K. The lowest temperature achieved was 77K, when the cooler was operated at maximum power for more than 30 minutes. These are very significant results. demonstrating the utility of low cost, readily available tactical cryocoolers for space experiments.

       Using the cryocooler as a vibration source, the response of the satellite was determined from accelerometer data. Results agreed with ground tests : forces of about 0.008g in the direction of the cryocooler expander at the fundamental frequency were measured. The satellite is relatively rigid at 55 Hz where the cryocooler is operated.

       In conclusion, the reliability and utility of tactical cryocoolers in space systems has been demonstrated. Although they are a source of high vibration levels (by spacecraft standards), the resulting motion can be controlled. Many of the lessons learned from the STRV-1b experiment are being applied in the design of a next- generation experiment to be conducted on STRV-2. Once again, a tactical cryocooler will be used, but in a more demanding role as the primary cooling for an active focal plane array in an infra-red optical system. Also, the digital electronics/software approach used for this experiment has proved very effective as noted, and is already appearing in vibration suppression hardware provided by cryocooler manufacturers.

       The objectives of this experiment were to demonstrate fault tolerant and graceful degradation characteristics of JPL's analogue neural network VLSI chips in a space environment, to correlate with ground-based radiation experiments and thereby lead to better device and circuit designs for space-borne neural network applications.

       Neural networks. based on massively parallel biological systems. perform efficient pattern recognition and optimization functions because of their capability to "learn" from examples. The architecture is particularly fault tolerant and lends itself to graceful degradation under harsh environments, rather than have a catastrophic failure of the chip. They are suitable for a variety of space, military, and civil applications and provide enhanced speed, autonomy, and intelligence. For this experiment, a special feedforward "learning" architecture (signal flow from one layer of neurons to next through a connecting mesh of synaptic weights) with a suitable pattern recognition problem as a bench- mark for learning was chosen. Reconfigurable synapses as multiplying digital-analog converters with weights stored in memory and neurons as variable gain amplifiers with sigmoidal output were used. Learning capability of the network by way of multiple identical VLSI fabricated chips was monitored on ground as in flight.

Ground-based results:

• Performance for unbiased chips degraded gradually from 0 to 100 kilorads (krads) when the neuron characteristics (sigmoidal curve) became steeper with dose. Chips were functional to 100 krads; beyond 100 krads the synapse current leakage and the memory errors increased and the chips failed at 150-400 krads.
• Biased chip performance degradation was more than an order of magnitude faster and the chips failed at about 5-7.5 krads. Monotonicity errors in synapse- neuron curves were predominantly responsible in gradual chip failure. Overall, high energy electron damage was more severe than that due to proton or gamma rays.

Space-based results:

       The RADMON instrument consists of 16 4-kbit SRAM's designed to detect protons, alpha particles and heavy ions (SEU-SRAM's). In addition. the instrument contains 4 p-FETs for detecting total ionising radiation dose (TD-FET's).

       The dose curves indicate the effectiveness of the shielding and also a 27 day cycle typical of sunspot rotation. This latter effect has also been observed by the STRV-Ia CREDO experiment and the STRV-1b REM experiment. The primary devices show an approximate dose rate of 100 rad/day (36,500 rad/year). This corresponds well with the mission design requirement of 20 krad. However, the shielded devices, which are more typical of components within the platform subsystems due to the amount of shielding, indicate a dose rate approximately an order of magnitude less than seen by the primary devices. However, the measured dose appears to be higher behind' thicker shields than predicted by existing radiation codes. A radiation gradient was noted across the JPL experiment box. where the dose 10cm from the edge of the box is 78% lower ban the dose at the edge of the box. Dose was successfully measured in spite of large temperature variations which ranged over 50° C.

       SEU-SRAM results show that upsets vary as expected with bin AVos and shielding, with the highest number of upsets recorded in the exposed proton bin. The number of upsets is highest at the beginning and end of data acquisition. where the satellite is passing though the proton belts.

       In ground radiation tests, it was found that these devices are resistant to proton irradiation: device barrier height degraded by only 0.25 meV/krad. The objective of the STRV-Ib experiment is to determine if the more complex spectrum of incident radiation might result in increased sensitivity on the part of the HIP sensor.

       REM is a simple system consisting of two independent shielded silicon detectors with different types of shielding. Energetic particles impacting on the detectors generate pulses which are counted, and the number impacting in a set time are stored. By separating pulses according to their magnitude (-E) which depends on energy, the particle type and energy data can be derived. Each detector has sixteen channels of discrimination on pulse magnitude. The detectors used are 300Tm silicon diode detectors. The "electron detector" is 25 mm2 with a 3 mm aluminium shield. while the larger "proton detector" of 150 mm: is shielded by 3 mm of aluminium and an additional 0.75 mm of tantalum. The extra tantalum considerably reduces the electron and bremsstrahlung penetration and so makes this detector better at monitoring protons. The electron detector shield stops protons of energies below 24MeV and has an electron cut off around 1.4MeV. The proton detector has a cut off energy of 35MeV and suppresses incident electrons by a factor -200. Despite the simplicity, by flying the experiment in GTO and operating the experiment over an extended period. with good time resolution, signif']cant data can be generated. Additionally, knowledge of the pointing direction can be used to investigate directionality in the environment.

       A copy of the REM is also mounted externally on MIR. Both REM models have been operating simultaneously during 1995. The correlation of the two instrument data sets will prove highly valuable in the context of modelling the environment. The results can also be correlated with those made by complementary experiments on the STRV-1 satellites and on other spacecraft. Calibration is important to verify the expected performance of the detectors and has been performed in particle accelerators, complemented by numerical simulations.

       The instrument is operated continuously around an orbit and under software control changes sampling rate depending on time from perigee. To date, REM has provided good quality data with few major problems. Initial problems with the on-board processing software were solved by uploading a modification to the software. Similarly, accumulation times have been optimised. Dead time corrections have also given some problems but this is felt to have been solved. Some noise at the most sensitive -E channels is also seen. Extraction of particle spectra has taken longer than expected due to the need to perform careful numerical simulations before applying it to the counts data. The counts data illustrate in qualitative terms the highly dynamic nature of the outer, electron belt and the relative stability of the proton belt. Correlations with the 27 day solar rotation period are seen implying long- lived solar structures well connected to the magnetosphere during this phase of the solar cycle (approaching solar minimum). This last year has seen several long-lived energetic electron enhancements reported from the NOAA GOES satellite and these shall be studied with the aid of STRV-1b data, which give good cross-sections of the belts during these periods.

       The REM instrument can also provide high time resolution dose data by summing the energy deposit spectra of the silicon detectors. These preliminary data imply that the environment is milder than expected, even though recent evidence shows that the electron environment is more severe at solar minimum.

       The experiment operates as expected in an 'untriggered', test mode, where IV curves are generated for each of the cells under test without reference to the relative position of the sun. The 'triggered mode' is reliant on two 'sun overhead detectors' which trigger the experiment to operate when the sun is within 5° of the normal to the panel. To date, this mode has not been made to operate correctly since there appears to have been a failure of one (or both) of the sun detectors or their interface with the experiment.

       Currently, the degradation of the GaAs generating panels (with 500tam coverglasses) is estimated to be only < 6%.