The light from the sidereal sky (stars, the Milky Way, bright nebulae and galaxies) is much brighter than the variable Thomson-scattered signal we wish to detect. Thus, these brightness contributions must be subtracted from orbit-to-orbit SMEI sky maps. The simplest way to remove this background is either by subtracting orbits adjacent in time (then only the heliospheric change in signal in adjacent orbits remains) or by subtracting an average sidereal orbit map that combines several adjacent orbits together. To first order, SMEI instrument changes and sidereal background changes combined are small enough that differences from the average over a few days do not degrade overall SMEI performance beyond the photometric limit. For time intervals longer than a few days, the Earth’s orbital motion around the Sun slowly re-orients the asymmetrical stellar images such that unresolved bright stars no longer subtract from one another properly and this adds significant noise to the resulting heliospheric map.
In the UCSD analysis an average Sidereal Map was formed from "Engineering Mode" data. These were recorded roughly bi-monthly for each camera with onboard binning switched off, this means that all pixels for a given camera are available for analysis. Due to telemetry limitations, data in this mode were returned from only one camera at a time. A typical calibration included one day's 14 to 15 orbits for each camera. For a given camera and each calibration period, a sky map was formed in RA and declination: each pixel in the map is the median of the individual-orbit-values after subtracting the parameterized zodiacal light. Use of the median reduces possible contamination from SAA and auroral-oval particles that survive formation of the individual-orbit sky maps; the median also reduces contamination from auroral light as it moves over the maps in the course of a day. During the first 6 years of SMEI data, 34 such daily maps were generated for each camera. The final sidereal residue shown here, which was subtracted within the main SMEI data analysis pipeline, was an average of the Camera 1 and 2 maps. Camera 3 data were not included in this average because this camera views much brighter zodiacal light, and also suffers a much higher noise level since it ran hotter than the others, and had many "flipper pixels".