Attending: Simon K., David C., Colin S., David N., Dominique B., Francisco, Hsin-Fang C., Jim B., Michael W.-V,. Robert L., Robert G., Paul P.

Notetaker: Simon

David N.: Working on COSMOS data and IDL wrapper, specifically with processCcdDecam.  It figures out list of visits and ccds, sends jobs to individual nodes, ascertains whether any fail by looking at logs and expected files.  Some basic QA as well: plots and web pages.  https://lavinia.as.arizona.edu/~dnidever/lsst/html/processCcdDecam_0176837.html

Lots of info about jobs including green for good red for bad, input configs, logs, processing time, etc.  Access to images.  You can see why things failed via logs.

Running all COSMOS next week (hopefully).  Doing makeCoaddTempExp next.  Getting error in reportImagesToCoadd (no column called tract).

Jim: We typically let the task loop.

David: I want finer grained control

Simon: I can take a look at reportImagesToCoadd if you send me the skymap.

Action: David send pointer to skymap for Simon

Hsin-Fang: Put location of the bias jump in the config.  Seems to work o.k.  Tried to run ISR and found new things to work on.  E.g. glowing edges are not masked (but they are in instcal).

Bad pixel mask for raw data is just for bad pixels (excludes edges).  Could bring them over from the community pipeline masks.

Robert L.: Surprised edges don’t flat field out.  We could take care of the edges as part of the bad pixel mask.

Robert G.: I think it may be that the non-linearity may be applied in different orders in the flat generation and flat fielding.  The size could drive some pixel values into the non-linear regime (this is a guess).

David N.: How do you deal with defects.

Robert L.: We have an ascii file format.  We can also use FITS file format.

Robert G.: Can we have defects that are suspect but not flagged as BAD?

Robert L.: We do something like this on the HSC side.

(aside from Simon) I don’t think we currently have a notion of BAD-BAD pxels and SUSPECT-BAD pixels, but we could certainly add this.

David: So how should we deal with this?

Robert: We should probably make the unusable edge pixels BAD in the flat and propagate them through.  We could also do this in the defect file stage.  We will probably throw some of the edge away and try to correct the less bad edge pixels.

Robert G: This is basically the approach we are using.  Mask totally the edge pixels and mask some pixels interior to that as SUSPECT.

Action: Hsin-Fang  will have  a look at defects files e.g. https://github.com/lsst/obs_subaru/tree/master/hsc/defects/2013-01-31

Action Simon: Look into no-linearity correction by LUT.

Colin:  Looked into warping with TPV.  Some catastrophic failures.  Wrote the code to pass TPV through.  Things were fine, but then persistence stripped out TPV coeffs, so diffim fails.  This is complicated because wcslib does some internal conversions, so we’d have to modify wcslib.  It’s a big hassle to write things out.

Robert L.: You can’t ask wcslib for a FITS header? If so, we can do this with cfitsio.

Colin: I’m not sure.  It may return some other format, but it may be doable.

Colin: My solution was to pass the TPV through, run approximateWcs and replace the TPV WCS with SIP.  That works, but it doesn’t fix any of the dipoles in diffim, so it may not be distortion terms.

Robert L.: DESDM does not use radial terms, so really this is just the same as SIP.  We are just fitting to too low order to fit out the atmosphere.

Colin: I think that’s my conclusion as well.

David N.: Are the distortion terms good?

Colin: We are trying to figure this out.  We are still getting astrometric dipoles, so we are not getting it “right.”

Michael: It could be the matching kernel not varying spatially correctly…

Robert G.: An additional issue is that the PV terms are fixed, so some of this could be coming from the chips moving around for various reasons.  We have noted that some CCDs move around more, but we have not propagated this.  The coadds are fit simultaneously, so this problem goes away in coaddition.

Robert L.: Simultaneous astrometry should probe this.

Dominique: Simon had been looking at running on DECAM

Simon: I never made it work for uninteresting reasons.  I can look into it again.

Action:  Dominique will modify afw code to read linear and constant terms in the SIP polynomials.

Action: Simon will look at running simultaneous astrometry on CP DECAM data.

David: DECALS team claim that they are getting really excellent astrometry out of DECAM.

Action: David will talk to Arjun Dey about DECALS astrometry.

Francisco:  Nothing major to report.

Dominique: Coaddition of Abell 2261 with no errors.  Restart work on simultaneous astrometry.  Check if coadd is improved with simultaneous astrometry vs using header WCS.

Robert L.:  We processed all our data through SFM on the mountain (on a macbook-pro).  We were sloppy about flats, but the ISR wasn't as good as we expected.  When Merlin gets back from DESY we'll sort this out.