AIRMASS inaccurate in MzLS file headers and ccd-annotated files

[djschlegel]

The AIRMASS values in the MzLS files aren’t particularly accurate. 10% of the files have them wrong by more than 0.05. They were less wrong during the first year of the survey, then more wrong after that (see attached plot vs. time).

An extreme example is this file (raw, CP): mosaic/MZLS_Raw/20170613/k4m_170614_043001_ori.fits.fz mosaic/MZLS_CP/CP20170613/k4m_170614_043001_ooi_zd_v1.fits.fz where the header lists AIRMASS=2.676, but the correct value would be 1.053.

I don’t think this matters very much. These airmass values aren’t used in the photometric calibrations (either the PS1 calibrations, or ubercal because Eddie recomputes the correct airmass). But it is used for the CCD_CUTS for deciding if exposures were shallower-than-expected.

[fvaldes]

The MOSIAC3 KTM uses the post readout telemetry packet for everything (because many things are not known before the readout) except for

RA, DEC, TCPHA, ZD, UTDATE, UT, MSEPEDF, FOCUS

Obviously hindsight tells us AIRMASS should have been in this list.

[djschlegel]

This scatter persists in the AIRMASS listed in the annotated-ccd files (copied over from the CP headers) and truth, with a standard deviation of ~0.05 and outliers that are just way wrong. I'd suggest we just recompute AIRMASS for all of the mosaic exposures in the annotated-ccd file using MJD_OBS + 0.5*EXPTIME/24/3600 as the timestamp. I've made a file with corrected airmass to which compare, using as the KPNO coordinates latitude = 31.9633, longitude = 360. - 111.600 : /global/project/projectdirs/desi/users/djschleg/dr9/ccds-annotated-mosaic-dr9-fixairmass.fits

[djschlegel]

Another detail: For the mosaic exposures, the timestamps in those headers are: DATE-OBS= '2018-02-12T03:12:11.0' / Date of observation start (UTC)
MJD-OBS = 58161.13346065 / MJD of observation start
MJDSTART= 58161.133486467 / MJD of observation start
MJDEND = 58161.135011267 / MJD of observation end

From what I could figure, MJDSTART may be closest to the shutter-open time. DATE-OBS and MJD-OBS have equivalent values in different formats, and are prior to that timestamp by 1.5-2 sec. MJDEND is approximately the shutter-closed time plus the readout time (~30 sec).

[moustakas]

We should just recompute every airmass value for MzLS, as suggested by @djschlegel.

[moustakas]

Code to recompute airmass was implemented in #470, so closing. @mlandriau will need to regenerate all the zeropoint files to incorporate this change into DR9.

[djschlegel]

Let's get these correct values into the ccd files. They're really pretty terribly wrong!