Data processing algorithms for on-board satellite event analysis

Event rates encountered by satellite-borne sensors typically exceed the available telemetry bandwidth, necessitating on-board data analysis. One method of reducing output telemetry requirements, employed by data processing units (DPUs) on several satellites (VIKING, CRRES, CAMMICE), is to partition the underlying event parameter spaces into disjoint classes, and to downlink statistics based on these classes rather than individual events. The DPUs utilize a hardware implemented binary search algorithm to classify each valid event into one of several mass and mass per charge-state (M/Q) groups, using look-up tables written under software control. Mass and M/Q group boundaries are defined via polynomial curves in the underlying parameter spaces. The boundary curves were obtained using a combination of global and local least-squares approximation algorithms applied to preflight sensor calibration data.<>