Regions of Bad Photometry » History » Version 10

Eli Rykoff, 10/29/2013 05:53 PM

1 1 Eli Rykoff
h1. Regions of Bad Photometry
2 1 Eli Rykoff
3 1 Eli Rykoff
-Eli Rykoff
4 1 Eli Rykoff
5 4 Eli Rykoff
h2. Bad Photometry Maps
6 4 Eli Rykoff
7 7 Eli Rykoff
Starting from the bad images below, I pixelized the sky and made a map of the fraction of objects in a given pixel (~0.5 deg^2) that has "crazy colors" in either g-r or i-z.  These colors are defined as g-r < -1 or g-r > 4 or i-z < -1 or i-z > 4 (using MAG_AUTO).  Of course there should be some number of faint objects that have these strange colors due to errors, but these should be negligible.  In fact, almost all of these objects are spurious detections of artifacts.  The following map shows the output map.  Note that the overall mask is chosen to have decent imaging in g,r,i,z according to a rough calculation of limiting magnitudes.  This is why the central strip at dec = -55 isn't covered, because this region is extremely shallow.  The red circles are stars brighter than R<5 in the Yale Bright Star Catalog.
8 4 Eli Rykoff
9 5 Eli Rykoff
10 4 Eli Rykoff
11 4 Eli Rykoff
While most of the survey is okay, some regions are not, with some pixels having as many as 25% of their objects with crazy colors.  We want to both mask these in SVA1 at the catalog level, and mask these in the individual images for Y1 runs.
12 4 Eli Rykoff
13 4 Eli Rykoff
h3. Streaks
14 4 Eli Rykoff
15 6 Eli Rykoff
Bright satellite streaks (which are yet unmasked) show up as long ~3 degree narrow clusters of bad objects in the coadds.  The good news is that these can be masked in the finalcut images because they are "unique" events doing this masking won't significantly decrease the depth in the final coadds.  The streaks range from bright satellites:
16 6 Eli Rykoff
17 6 Eli Rykoff
18 6 Eli Rykoff
19 6 Eli Rykoff
To crazy insane streaks.  The following (at RA~65, Dec~-57) appears to be an airplane.  Looking at the raw images (before the sky subtraction screws things up) it's clear that this is low enough to be out of there's an extra blinking light.  It should be possible to use the size of the streak to calculate the altitude of the airplane...
20 6 Eli Rykoff
21 1 Eli Rykoff
22 7 Eli Rykoff
23 7 Eli Rykoff
h3. Scattered Light
24 7 Eli Rykoff
25 7 Eli Rykoff
Considerably more troublesome is the scattered light from bright stars.  In particular, there is a large patch at RA~70, Dec~-63 that is completely compromised.  In these region is R Doradus, a bright (R~5) Mira, which means it is very, very red which may be why it's causing us such fits:
26 7 Eli Rykoff
27 7 Eli Rykoff
28 7 Eli Rykoff
29 10 Eli Rykoff
Note that if you look at individual chips where the scaling is more appropriate (my DS9 fu fails me with the mosaics) then the scattered light contaminates basically the full image).
30 10 Eli Rykoff
31 7 Eli Rykoff
However, the scale of the problem is surprising.  This is more than a degree away from the corner, and the scattered light may also be from R Doradus:
32 7 Eli Rykoff
33 8 Eli Rykoff
34 7 Eli Rykoff
35 7 Eli Rykoff
I'm not sure how to mitigate this going forward.  At the moment, we may need to cut a large patch out of the corner of the SPTE region because the photometry isn't reliable.
36 7 Eli Rykoff
37 1 Eli Rykoff
h2. Background Subtraction Issues
38 8 Eli Rykoff
39 8 Eli Rykoff
One more thing!  As has been noted in Peter & Erin's visual inspection tool, sometimes SExtractor has trouble with the background model in the corner of chips.  This may be more widespread than I (at least) thought was a problem.  Here is the background mosaic for one of the satellite images:
40 8 Eli Rykoff
41 8 Eli Rykoff
42 8 Eli Rykoff
43 8 Eli Rykoff
I've included the cutout graphs as well.  You can see that (a) three of the chips have bad background subtraction in the corner and (b) the background is somewhat concave rather than convex.  Presumably this is because of flat field issues...
44 8 Eli Rykoff
45 8 Eli Rykoff
In addition, you can see the amplifier shift on CCD 31, which is presumably being fixed.
46 8 Eli Rykoff
47 8 Eli Rykoff
My only thought on the bad corner background model problem is that a "global" fit to the background wouldn't have this problem since we know the background should be roughly contiguous across chips.  Though this may not be trivial to implement.
48 6 Eli Rykoff
49 9 Eli Rykoff
There are also small jumps from CCD to CCD in the background model.  What is the cause of this?
50 4 Eli Rykoff
51 4 Eli Rykoff
h1. Original Post
52 4 Eli Rykoff
53 1 Eli Rykoff
Brad Benson looked at SPT S/N for all the redMaPPer (v5.7) clusters in the SPT region.  He noticed there was a class of high richness clusters with no SPT signal.  What's up with that?
54 1 Eli Rykoff
55 1 Eli Rykoff
h2. The bad clusters
56 1 Eli Rykoff
57 3 Eli Rykoff
These are riz color composite images of the bad clusters.  Note that all of them are junk regions due to scattered light!
58 2 Eli Rykoff
59 2 Eli Rykoff
print, j, spt_sn(ind(j)), data(ind(j)).zred, data(ind(j)).lambda_zred, data(ind(j)).ra, data(ind(j)).dec 
60 1 Eli Rykoff
       1     -1.85195     0.707742      315.477       69.466146 -61.502703
61 1 Eli Rykoff
62 1 Eli Rykoff
       2     0.839335     0.743467      237.402       69.810693 -61.506111
63 1 Eli Rykoff
64 1 Eli Rykoff
       3     -1.07636     0.807284      226.846       14.894564 -49.750827
65 1 Eli Rykoff
66 1 Eli Rykoff
       5     0.223801     0.888624      192.095       15.071309 -49.751638
67 1 Eli Rykoff
68 1 Eli Rykoff
       6     0.172056     0.808576      191.317       69.758349 -62.785692
69 1 Eli Rykoff
70 1 Eli Rykoff
       7    -0.903793     0.730482      167.473       75.576196 -49.709253
71 1 Eli Rykoff