Order Tracing

By default, orders are traced with a density clustering algorithm (sklearn.cluster.DBSCAN) on either the flat fields (preferred) or data, but hard-coded maps may be used if the order locations on the detector are known to be relatively stable. If order locations must be uniquely determined from the data, tweaking the dbscan algorithm will likely be necessary. The precise order locations may still be refined for a unique exposure via iteratively cross-correlating an estimated trace profile (seeing profile) with the unrectified 2d image, so estimated order map algorithms are both sufficient and preferred, except in the case of crowded orders.

Calibration

Flat, bias, and dark calibration are performed when provided and set. Telluric calibration (flat star observations) Wavelength calibartion via ThAr lamps or LFC’s are not currently supported, but intended to be in the future if enough desired. Wavelength telluric calibration would be performed post-extraction.

Extraction

The trace profile (seeing profile) is estimated by rectifying the order and taking a median crunch in the spectral direction on a high resolution grid (tuneable parameter). The background sky, sky(x) is computed by considering regions of low flux (< 5 percent) within a given column. By default, an optimal extraction is iteratively performed on the non-rectified data, although the trace profile is interpolated for each column via cubic splines from it’s pre-defined fiducial grid according to the trace positions. Depending on the nature of the user’s work, this may not be suitable and one should rely on using an instrument specific reduction package or implementing one’s own optimal extraction algorithm(s).

Support Status:

1. iSHELL / IRTF (Kgas, K2, J2 modes via flat field order tracing)

2. CHIRON / SMARTS 1.5 m (highres mode, R~136k, under development)

3. NIRSPEC / Keck (K band, under development)