On board science data processing for NGST: algorithms and architectures

In the frame of an ESA program, as a potential contribute to NASA-ESA NGST project, this study focuses on the On Board Data Management System, starting from the requirements defined by a European Science Team. The expected amount of data collected in a single day of observation will be about 1600 GBytes, while the downlink capacity is at present in the range of 5 to 40 GBytes/day. Given these figures, on-board data reduction will be mandatory. It should include detector and signal conditioning, non-linearity correction, cosmic rays detection and rejection, source intensity (photon flux) estimation and data compression. The paper emphasizes different data reduction algorithms and compares their performances with respect to cosmic ray detection efficiency and source intensity estimation error. Monte Carlo simulations have been run on a range of cosmic ray hit probabilities, of source intensities and of readout noise variance. The most representative algorithms from the computation point of view have been implemented onto the most likely candidate processors for the NGST mission. Results are presented in terms of on-board computer complexity requirements, in particular number of processors and memory needs. Possible Science Data Processing architectural concepts are analyzed in order to come to a preliminary design and performances and budgets estimation; a solution based on an advanced use of Serial DS-Links IEEE-1355 network, whose main characteristics are scalability and technological independence. The baseline architecture is presented, highlighting the critical areas and the required technology roadmap.