2022AJ....163..130R

Neutron Star Interior Composition Explorer has a comparatively low background rate, but it is highly variable, and its spectrum must be predicted using measurements unaffected by the science target. We describe an empirical, three-parameter model based on observations of seven pointing directions that are void of detectable sources. Two model parameters track different types of background events, while the third is used to predict a low-energy excess tied to observations conducted in sunlight. An examination of 3556 good time intervals (GTIs), averaging 570 s, yields a median rate (0.4-12 keV; 50 detectors) of 0.87 c s^–1, but in 5% (1%) of cases, the rate exceeds 10 (300) c s^–1. Model residuals persist at 20%-30% of the initial rate for the brightest GTIs, implying one or more missing model parameters. Filtering criteria are given to flag GTIs likely to have unsatisfactory background predictions. With such filtering, we estimate a detection limit, 1.20 c s^–1 (3σ, single GTI) at 0.4-12 keV, equivalent to 3.6 x 10^–12 erg cm^–2 s^–1 for a Crab-like spectrum. The corresponding limit for soft X-ray sources is 0.51 c s^–1 at 0.3-2.0 keV, or 4.3 x 10^–13 erg cm^–2 s^–1 for a 100 eV blackbody. These limits would be four times lower if exploratory GTIs accumulate 10 ks of data after filtering at the level prescribed for faint sources. Such filtering selects background GTIs 85% of the time. An application of the model to a 1 s timescale makes it possible to distinguish source flares from possible surges in the background.