Post-processing modeling and removal of background noise in space-based time-of-flight sensors

This paper develops and implements a mathematical framework that enables noise modeling and removal for any sensor that relies on discretely measured events. Here, we apply this technique to data from the Fast Imaging Plasma Spectrometer (FIPS), a time-of-flight mass spectrometer on the MESSENGER spacecraft. An iterative Monte Carlo event-processing algorithm is used to probabilistically separate instrument measurements into real and noise-based events. Kernel density estimation is employed as a smoothing technique to enable noise removal for datasets comprised of only a few events. Given an accurate noise model, the overall misidentification of events is expected to be less than 25% even for datasets having low signal-to-noise (SNR) ratios, with substantially improved results expected for progressively larger accumulations of data. These techniques are shown to successfully recover heavy ion events from in-flight FIPS data both inside and outside Mercury’s magnetosphere. Such data analysis methods not only drive a more in-depth understanding of sensor operation, but also provide a unique post-processing approach that can result in the improvement of in-flight SNR without any modifications to instrument settings. In addition, these method are readily applicable to existing archived datasets of past missions.