Background subtraction is performed whenever there are blank-sky data available and the background subtracted data cube is made available within the GINGALAC and GINGABGD databases. Occasionally, we rejected data acquired during periods when the sun angle between the satellite pointing direction and the Sun was less than 90 degrees or the high voltage supply was reduced to V.
The background is estimated by modelling the known contributions to the count rates and periodicities observed in them. Both the local and universal methods described in Hayashida et al. (1989) and Williams et al. (1992) are used (the background subtraction method is written to each record in the database table). The background count rate, , dependent on both energy, E, and time, t, is given as
The first term represents the CXB contribution, which is constant in time, but varies across the sky due to fluctuations in the CXB. The remaining, time-variable instrumental components are characterised by their spectra F and associated HK parameters P (e.g. COR, SUD). The spectral form, F , was determined by fitting data from background observations as described below.
In modelling the component of the background due to the radioactive decays, we calculated a time counter, accumulated with an exponential decay factor, after each passage through the SAA. In practice, we used up to four time counters with exponential decay timescales corresponding to half-lives of 8 hours, 41 minutes and 20 minutes (see Hayashida et al. 1989 and Nandra 1991). The number of decays required in the model was reduced during the latter half of the mission to only one component, because the decay of the Ginga orbit caused the LAC to be further shielded from the Earth's radiation belts. The universal model uses all background observations within a contemporary three to four month period to model systematic trends in the particle levels, and hence to estimate the background level at the time of source observation. A specific time counter is required to account for the 37-day precessional period of Ginga. The local method was used exclusively during the first six months and last three months of operation of Ginga as the gradual rise and fall of the background made the universal method unreliable.
The uncertainty in the background subtraction is limited by source confusion and systematic errors rather than statistical uncertainties in the data. The limiting senstivity for source detection was chosen to be 2.1 counts sec-1, equivalent to a detection (Hayashida et al. 1989). All observations whose background subtracted 2-10 keV top-layer count rate exceeded 2.1 counts sec-1 were removed from the background modelling pipeline. This can be achieved only after background subtraction, and so the whole process is an iterative proceedure. The LAC mid-layer is insenstive to X-ray photons below keV (Turner et al. 1989). Therefore, all observations were rejected from the background modelling pipeline if the background subtracted mid-layer count rate below 6 keV was significantly above or below zero. In addition, observations were rejected from the background modelling pipeline if the background subtracted spectra and light curves (3-10 keV band) showed trends with energy and/or time.