Once your object traces are defined, click the "Extract" button to produce 1D traces of the object and standard spectra.  These will be in units of non-flux-calibrated counts.

There are three options for how extraction is performed:

Optimal Extraction

For optimal spectral extraction, an iterative, local background model is fit to improve estimates of the sky flux for subtraction.  This process takes a minute or so, and you will see information about the bspline fit scrolling to the terminal and also a plot of the trace cross section and fit, collapsed along the wavelength direction.  When the background model is finished, a 2D plot of the fully sky subtracted spectrum will be presented for inspection.

On approval, the code extracts the spectrum, using the calculated model object profile to perform optimal weighting for signal-to-noise (a la Horne et al).

This mode will improve the SNR of faint-object spectra, but provides little gain for bright objects dominated by photon shot noise from the object.  For those cases the flux estimate could be biased by imperfect modeling of the object profile, so for very bright sources we recommend Boxcar Extraction.

Boxcar Extraction (with local sky subtraction)

In this mode, the iterative improvement is made on the sky background model, which significantly improves upon the first pass sky subtraction done in the object finding stage.

However, for the final extraction the flux is estimated simply by summing up all (sky subtracted) pixels in the extraction aperture.  No weighting function is applied.

Boxcar Extraction (no local sky subtraction)

In this mode, the sky model consists of the first-pass estimate calculated during object finding.  No local correction is applied.  This method is significantly faster but suffers from larger residuals particularly around OH sky lines.  It may be appropriate for objects that are much brighter than the sky (i.e. standards) or when reduction speed is critical (e.g. at the telescope).

When should I use Optimal vs Boxcar?

These are our recommendations.  As usual YMMV.

  • Use optimal when your object is faint, meaning anything comparable to the sky brightness or smaller
  • Use boxcar (with local sky subtraction) for almost anything else, especially when your object trace has ~10000 counts or more
  • Use boxcar (without local sky subtraction) if you can't wait the ~1 minute it takes for local sky subtraction, or if your object is close to saturation.
  • No labels