We digested our Twinkles 1 results and identified some action items. Then we looked at our Twinkles 2 Plans, and identified issues to resolve and features to add.
Discussions:
Phil:
Center of field …not exposed to flat fielding problems
Big Dithers ?
Should DC2 be combined Twinkles for PhoSim Deep analogs ?
Things like flat fielding will be done in PhoSim Deep. So focussing on a chip without these problems is important
Tom :
Why did we turn off Clouds? : not turned on because the variation seemed large and wanted to start simple. A Also at that point we were not doing spatially varying zeropoints which is being worked on now.
Long jobs: (these were 2109 jobs that were predicted to take too long to run based on a model of Run 1.1.)
Is it ok to not have these nights in the analysis? Probably … will say that we have explored parameter regions.
Twinkles Metadata:
We should produce a catalog/database of all of the single-number parameters characterizing PhoSim runs so that we can diagnose why long runs fail.
Use cases:
Correlate execution time with moon parameters
Correlate processCCD execution time with PhoSim execution time.
Correlate processCCD execution time with number of input sources
Correlate processCCD execution time with PhoSim seeing and DM measured seeing
Correlate DM failures with properties of PhoSim runs
What are the last 10 lines of the failure. What’s the link to the full output.
How many
These all should be accessible with a simple “one-line” query call suitable for implementation by a variety of quick scripts. This information should be easily available to DESC members with no more than 2 minutes required to look up the information and no external help or additional authorization required beyond having a NERSC account.
Parameters to track:
Number of sources in InstanceCatalog
PhoSim header entries
brightest source(?)
Should we have this for other projects too? Probably yes
CModel fluxes is on the basis of bulge or disk (using best fit)
Action Item: Turn on moments
Visit images : the images thrown out are less than 1 percent of the images. < 0.6 seeing. What is the median ?
How can we get the workflow tool on a local machine? Why ? For example there are loops in the DM functions that won’t be available to a workflow…
Decision: difference images against DR1 templates only.
Possible problem: there do not appear to be u-band images in year 1
Issues:
Features:
Actions: