This particular recipe is used for quality control and instrument monitoring.

The input list of frames is browsed in the order they are received. For every pair of consecutive frames that share the same DIT, NDIT, and readout mode, the readout noise is computed. Notice that the recipe does not try to make all possible pairs from the input list, so it is important to make sure that the files are correctly sorted. If the frames are kept in the order they were acquired, this should not be a problem.

For each pair, the second frame is subtracted from the first. The resulting image is then cut out into 4 quadrants and on each quadrant the following measurement is applied:

- Generate 100 9x9 windows on the input pixel surface. These windows are optimally scattered using a Poisson distribution to make sure they sample the whole area with as little overlap as possible.
- Compute the pixel standard deviation in each window.
- The readout noise is the median of all these measured standard deviations.
- The error on the readout noise is the standard deviation of all measured standard deviations.

The results are written into a PAF file (readable ASCII). One value per quadrant is printed out.

Notice that the algorithm is identical in LW mode, but the whole detector is considered instead of cutting it into four quadrants.