Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: pipe_analysis
-
Labels:
-
Story Points:6
-
Epic Link:
-
Sprint:DRP S21a (Dec Jan)
-
Team:Data Release Production
-
Urgent?:No
Description
The latest RC2 run was run with w_2020_50 (DM-28140). This is the first time we are running with the gaia_dr2_20200414 reference catalog for the external astrometric calibration in jointcal, so some small differences in the astrometry would be expected. However, the differences are not small. To start, the stellar locus is now of order 33 mmag, whereas in the v21_0_0_rc1 run it was at 13 mmag. An "issue" is clearly visible in the stellar locus plots:
vs.
Also striking are the patterns seen in a match to the reference match catalog, e.g.:
Note that the plotting scripts are still loading in the ps1_pv3_3pi_20170110 to do the matching (to be updated on a forthcoming ticket), but initial investigations confirm that this cannot explain the amplitude & shape of these patterns.
This ticket is to look into any possible cause of these issues.
Attachments
Issue Links
Activity
To follow up on a point raised by Ian Sullivan on our Jan 4 Metrics/QA meeting, one of the changes that went into the w_2020_50 was the issue of DM-27868 (effectively adjusting the center of the ref cat search radius). A check on the effects of this change is possible by using the doApplyExternalSkyWcs[1/2]=False config option in the visit-level pipe_analysis scripts (where the [1/2] are used in the compareVisitAnalysis.py script for the comparison reruns...both were set to False for this test). I ran this for both the v21 and w50 runs and the results are indeed nearly identical (and if anything, the newer run is slightly better in terms of mean offset/scatter). Here are some demonstrative plots:
w50:
v21:
compare v21 vs w50:
So I think all is good here (note the scales and metric values!)
Do you think we're good to close this out now? I've updated the story points to reflect the time I spent on this (half of which was just convincing myself this wasn't "my fault"!), but I would think at least [~erykoff] should claim some SPs for his efforts (you too!)...perhaps he could absorb the appropriate number into DM-28233 (if he hasn't already done so)?
I'm going to attach this for posterity: https://github.com/yalsayyad/dm_notebooks/blob/master/calibration/Jointcal%20with%20Gaia%20RefCat.ipynb and add 2 SPs that I spent on this too.
Awesome...thanks!
Excellent. That's much better than the referenceErrors=10mas bandaid: https://github.com/yalsayyad/dm_notebooks/blob/master/calibration/Jointcal%20with%20Gaia%20RefCat.ipynb
We need something for w02 which gets tagged tonight. I put an item on the agenda for team meeting.