Charge Transfer Inefficiency Correction

Overview

Charge traps are found in the serial registers of several LBNL CCD quadrants. In the DESI pipeline, two algorithms have been developed to mitigate the effect.

CTE correction during preprocessing

A model for the effect of charge traps has been developed. For amplifiers with keywords CCDCOLSX=BEGIN:END with X being an amplifier id (A,B,C,D) in their calibration (as found with the CalibFinder with looks into yaml file in the $DESI_SPECTRO_REDUX repository), a model is fit on flatfield LED exposures and saved in tables $DESI_SPECTRO_CALIB/$SPECPROD/calibnight/NIGHT/ctecorr-CAMERA-NIGHT.csv

For the same cameras with the keyword CCDCOLSX, the model is applied during preprocessing to correct the CCD image. It relies on an iterative fit where the spectra are extracted from the CCD image, then projected back to obtain a CCD image model that is used to compute the effect. The effect is then subtracted from the true image.

This is performed ONLY for CCD amps with the keyword CCDCOLSX

CTE correction during sky subtraction

For CCDs with the keywords OFFCOLSX=BEGIN:END OR CCDCOLSX=BEGIN:END, a offset and a slope as a function of wavelength are fit at the same time as the sky model for fiber traces overlapping the affected region (defined by the CCD columns BEGIN to END). This method is sufficient for charge traps that hold just a few electrons but is insufficient for traps with larger effects. For those using the keyword CCDCOLSX also triggers a correction at the preprocessing level.