William M. Johnson

Space avionics stellar-inertial subsystem

Draper Laboratory has developed a space avionics system for attitude determination and control of small low-power microsatellites. The goal is to determine microsatellite attitude to better than 0.1 deg with a minimum power expenditure. The revolutionary, enabling technologies that make this happen include an electron-bombarded Charge-Coupled Device (EBCCD) star camera that can detect dim stars (Magnitude=8) and a MEMS 3-axis inertial rate sensor module. The performance of the two enabling sensors (EBCCD and MEMS) depends on the early amplification and digital conversion of the optical and inertial signals, respectively, to a digital format. The amplification of the optical signal in the EBCCD star camera removes the effect of the CCD detector noise and the effect of the read-out noise. The digital conversion of the inertial sensor data at the preamplifier stage eliminates the subsequent effect of the noise and drift error terms. The immediate NASA application is microsatellites in LEO NASA science missions that often require 20 rpm spinning microsatellites. This paper describes some of the design features of the Draper Low-Power Avionics Sensor Suite (LoPASS) approach, including the digital processing of the EBCCD and MEMS data. Examples of the LoPASS sensors digital processing include: use of the UTMC 69R000 microcontroller to implement the digital processing; measuring star centroids on a discrete CCD array with high accuracy; providing digital temperature compensation for the MEMS gyros; application of a simple Kalman filter to optimize the attitude solution; low-power operation with radhard sensors and processor. The result is a space avionics stellar-inertial subsystem that weighs approximately 1 kg and has an average power of approximately 2 W. The optical camera, inertial sensors, and the microcontroller are all radhard. The star camera is turned on every 5 min with a 3-second star sighting to update the MEMS gyro drift and achieve the 0.1-deg attitude knowledge requirement.

Algebraic Models For Detection Probability And False Alarm Rate

The problem of characterizing detection system performance presents the following dilemma: On the one hand, if the system is good, then performance failures will be extremely rare events. But on the other hand, if the occurrence of rare events is to be characterized, then reliance entirely on Monte Carlo simulation would require an enormous number of runs, and the expense would be prohibitive. Algebraic models contribute a complementary approach to circumvent the problem. Such models can augment the simulation approach in a variety of ways. This paper discusses several such models, focussing particularly on models for detection probability and false alarm rate. Numerical results confirming specific models are presented.

Accurate Centroid Tracking

The object of this work was to develop analytical tools for, describing errors associated with mosaic array centroiding. As an example of a current problem, a Monte Carlo simulation of a point source tracking problem was implemented. Then for a given noise and image model, the accuracy of the image tracking was evaluated. In general the various causes for error in the centroid estimate were pixel geometry, device and background noise, and energy spillage outside window. Both mean errors (bias) and distribution about the mean (variance) were studied. The conclusions were: Various sources of centroid estimation errors were developed. A Monte Carlo simulation of track position estimation was evaluated for various scenarios. Multiple measurements could be shown to provide resolution of a fraction of a pixel.