# Description¶

## Assumptions¶

It is assumed that the input science data have NOT had the zero group (or bias) subtracted. We also do not want the dark subtraction process to remove the bias signal from the science exposure, therefore the dark reference data should have their own group zero subtracted from all groups. This means that group zero of the dark reference data will effectively be zero-valued.

## Algorithm¶

The dark current step removes dark current from a JWST exposure by subtracting dark current data stored in a dark reference file.

The current implementation uses dark reference files that have been constructed from exposures using nframes=1 and groupgap=0 (i.e. one frame per group and no dropped frames) and the maximum number of frames allowed for an integration. If the science exposure that’s being processed also used nframes=1 and groupgap=0, then the dark reference file data are directly subtracted frame-by-frame from the science exposure.

If the science exposure used nframes>1 or groupgap>0, the dark reference file data are reconstructed internally to match the frame averaging and groupgap settings of the science exposure. The reconstructed dark data are constructed by averaging nframes adjacent dark frames and skipping groupgap intervening frames.

The frame-averaged dark is constructed using the following scheme:

• SCI arrays are computed as the mean of the original dark SCI arrays
• ERR arrays are computed as the uncertainty of the mean, using \frac{\sqrt {\sum \mathrm{ERR}^2}}{nframes}

For each integration in the input science exposure, the averaged dark data are then subtracted, group-by-group, from the science exposure groups, as follows:

• Each SCI group of the dark data are subtracted from the corresponding SCI group of the science data
• The ERR arrays of the science data are not modified

Any pixel values in the dark reference data that are set to NaN will have their values reset to zero before being subtracted from the science data, which will effectively skip the dark subtraction operation for those pixels.

The dark DQ array is combined with the science exposure PIXELDQ array using a bitwise OR operation.

Note: If the input science exposure contains more frames than the available dark reference file, no dark subtraction will be applied and the input data will be returned unchanged.

## Subarrays¶

It is assumed that dark current will be subarray-dependent, therefore this step makes no attempt to extract subarrays from the dark reference file to match input subarrays. It instead relies on the presence of matching subarray dark reference files in CRDS.