Computing the run pointing in sky coordinates

[moralejo]

Addresses #1293
So far the first event's pointing was used. This is problematic for runs in which the pointing is unstable at the beginning.

[moralejo]

Possible improvements:

• check stability of pointing (and at least display a warning)
• return the "first reliable event" and use it to omit the previous ones from the DL3 file?

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 73.50%. Comparing base (1d20811) to head (9df60c6).
Report is 15 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1302      +/-   ##
==========================================
+ Coverage   73.48%   73.50%   +0.01%     
==========================================
  Files         134      134              
  Lines       14201    14211      +10     
==========================================
+ Hits        10436    10446      +10     
  Misses       3765     3765              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[moralejo]

I hope it is ok to do the change in docs/conf.py in this PR
That was just a deprecated option which made docs building fail.

[mexanick]

Per astropy docs for AltAz, the optional parameter pressure defaults to 0, effectively turning off the refraction. This will give about 300 arcsec bias in the transformation.

[maxnoe]

Not for Cherenkov light produced in 10 to 20 km height

[moralejo]

Here we want to know where the gamma-ray source "really is", ¿right?, rather than where its optical image (if any) would appear on the camera. Any bias in g-ray direction reconstruction related to refraction must be addressed via simulations (i.e. by training algorithms with MC which implements C-light refraction) .

This said, I had not even thought about that... so good that you ask (and good that the default is P=0)

[moralejo]

I hope it is ok to do the change in docs/conf.py in this PR That was just a deprecated option which made docs building fail.

Done now in #1303

[morcuended]

Something is not okay with the downloading of the test dataset in the CI. It's taking too long

[moralejo]

One unrelated test failed. Trying to fix it in #1307 #1300

[morcuended]

Looks good to me

[maxnoe]

This compares the median over all events to the median over the events without the ones excluded by "fraction".

Is this really what we want? Normal runs are long (~20 min) and the unstable pointing we saw was for the first ~35 seconds in the bad cases.

So the median won't be affected by the the first 35 seconds of unstable pointing and then stable pointing.

We looked at the deviation of the individual pointings from the "good pointing" in chunks of 2 seconds.

[mexanick]

If we want to detect unstable outlier values at the beginning, why don't we just use a sigma clipping routine from scipy or astropy?

[moralejo]

This compares the median over all events to the median over the events without the ones excluded by "fraction".

Is this really what we want? Normal runs are long (~20 min) and the unstable pointing we saw was for the first ~35 seconds in the bad cases.

So the median won't be affected by the the first 35 seconds of unstable pointing and then stable pointing.

We looked at the deviation of the individual pointings from the "good pointing" in chunks of 2 seconds.

Right, this will not be sensitive to the instabilities if they are so small... Perhaps it makes more sense to simply write out in the log what fraction of the run has a pointing beyond, say, 0.05 deg of the target. But I do not know where it would make sense to set a threshold for a warning.

[moralejo]

If we want to detect unstable outlier values at the beginning, why don't we just use a sigma clipping routine from scipy or astropy?

For simplicity, we know the problem occurs only at the beginning of the run. At least the problem we are addressing in this PR. And sigma clipping requires some settings, not obvious to me which ones would be adequate here.

[morcuended]

It looks good to me as it is. Just stress @maxnoe comment #1293 (comment):

Maybe we should identify the time range of this unstable pointing and exclude it via the GTI.

So in the long run we want to exclude completely from the analysis those first seconds of the runs via GTI definitions (if there is unstable pointing)