Previous version of this page can now be found here: DetectorLinearityArchive

Detector Linearity

The final linearity correction is actually more of a flux dependent bias level. The method used to generate the correction data was:

1. Run ppImage on input flat images with nothing but the OVERSCAN correction enabled.
2. Measure cell and edge row/column median values.
3. Foreach chip/cell/region/observation date, find the best fit trend f_region = gain * t_exp + bias_0 over a range of exposure times 3 < t_exp < 12.
4. Define the expected model value as MODEL = gain * t_exp and correction CORR = f_region - (gain * t_exp). This ensures that the bias offset bias_0 is included in the correction.
5. Foreach exptime/chip/cell/region, take the median of all correction values and fluxes, and write the table of (flux,correction) pairs.
6. Apply the correction by linear interpolation of the pairs for the appropriate chip/cell/region.

I have attached two images that show the original overscan corrected image and the linearity corrected version of this image. These two images are from 2009-12-30 and 2010-11-23, showing that the correction is stable.