Update flat threshold in comp_make_flat

[mgalloy]

The calculation of the minimum threshold for the median of flats is currently computed via line 337 of comp_make_flat.pro:

threshold = (norm - min_flat_median_offset) * expose / 250.0 / transmission_correction

where norm is the transmission computed for the day with comp_transmission and min_flat_median_offset is an epoch value that is currently always 20.0.

But the median of the flat image has been normalized already, so norm - min_flat_median_offset shouldn't be a function of date. So the threshold should be computed:

threshold = min_flat_median * expose / 250.0 / transmission_correction

where min_flat_median_offset has a constant value. We need to find this value through analysis of the flat medians.

Note, by 20110622, there are not as many flats currently marked as bad, and the bad ones are just barely bad, from the logs from 20110622:

  38.60 (flat median) < 38.86 (minimum theshold)
  38.82 (flat median) < 38.86 (minimum theshold)
  38.85 (flat median) < 38.86 (minimum theshold)

Tasks

• update equation for calculation of minimum threshold for flats — min_flat_median should 70.0 before 2016 and 60.0 after
• analyze statistical properties of the normalized (corrected by comp_transmission) median flat values — they should be roughly constant over the mission
• determine new min_flat_median_offset value (min_flat_median of 70 before 2016, 60 after 2016)
• create all flats for the mission with both old and new criteria, and warn when criteria disagree (log filename, median value of the flat, median of flat before converting to millionths, and both criteria thresholds, and maybe median(flat - dark) if possible)

[mgalloy]

Below is the most recent flat medians plot.

flat-medians-20201222.pdf

[mgalloy]

Before 2016, 70.0 should be minimum transmission corrected values. After 2016, 60.0 should be good.

[mgalloy]

I have the code ready to create all the darks/flats again and log differences in the passing by the old and new thresholds. I will need to untar each month of raw data as the processing progresses.