dispersion_calibarion

Dispersion Correction

As far as I can see, the best way of wavelength solution for now is using the sky OH lines, which requires decent exposure of empty sky of the night. And the wavelength calibration is currently done in two steps.

1. First, we use ThAr lamp image to guess the initial wavelength solution. This requires a reference wavelength solution (from a previous observations), and the reference solution is transformed to match the reference ThAr image to the newly obtained ThAr image. Currently, only an affine transform is made (shift + rotation + skew). Also, the reference solution is based on March observation. And replacing K band VPH grating after march run changed the solution quite a bit, and we get relatively large error (~ 1 pixel) for K (~0.5 pixel for H).

2. Second, we use sky lines to do full wavelength calibration. We start from the initial solution from ThAr line which should be better than 1 pixel. From that, we fit the pre-defined list of sky lines (~200 lines per band) to derive the wavelength solution. For each band, the wavelength solution is derived by fitting with 2-d Chebyshev polynomial.

For H band, we only use OH lines.

For K band, there is little OH line in the longer wavelength orders. For this orders, we use lines from hitrans spectra obtained for CRIRES (http://www.eso.org/observing/dfo/quality/CRIRES/pipeline/pipe_calib.html). Note that these lines are broad and not as good as OH lines.

The two step process is designed so that the pipeline can be run on the fly during the observing run. Before the observation begins, ThAr lamp is observed which will give initial wavelength solution. During the night, empty sky will be observed which will give full wavelength solution for the night.

My assumption here is that wavelength solution do not change throughout the night.

In the future, when telluric correction is correctly handled by the pipeline (i.e., fitting the telluric lines with atmosphere mode), I plan to incorporate a fine tuning of the wavelength calibration as a part of telluric correction. This will results in a fine-tuned wavelength solution per a target.

Distortion correction

We also use sky lines to correct the shift of wavelength solution along the slit position. The amount of shift is currently about 1 pixel. Unfortunately, Both ThAr lamp and UNe lamp has artifact (whose reason I do not know), which prevents me from using these lamp.

1. Sky spectra are extracted from different positions along the slit, and wavelength solution for each slit position is derived (2-d Chebyshev). This gives a set of wavelength solution (given o and x ( order and pixel position ) returns lambda).
2. For each order, wavelength solutions from different slit positions are fitted again to produce a distortion function. The distortion function describes the amount of shift in pixel (along the dispersion direction, where the reference is wavelength solution from the slit center) as a function of slit position and x.