Very Faint J mag Sources

Rae sent me these sources with very faint j mags:

1) 2.891835 40.401436 j_m = 19.5, j_msig = 6.788 date 981101n, scan 055, ID 9061 (cntr 226203530), J & H bands only, the K band src is very close, ID 9059
2) 0.827970 38.241749 j_m = 22.2, j_msig = 9.999 date 981101n, scan 035, ID 4528 (cntr 226034525), J & K bands only

Source 1:

Apparently this faint source didn't completely bandmerge; it only has J and H band data, and there's a K-band-only source very close to it. The high background noise levels probably shifted slightly the source's positions in one or more bands. Here is the pertinent s055.cal file data for the J & H source, the K-only source, and the "best" (saturated in Read1) extraction of the bright source nearby, split into 4 lines each (it's best to view this with a wide window):

| IDNum|    RA    |    Dec   |EMaj|EMin|Ang| X_u |   Y_u   | AssnCatID | R | Th| Mag1| Mag2|O|I|CnfFlg|
   9061   2.891835 40.401436 0.51 0.51 90   80.83 -5937.23 U127500122110 0.28 131 19.00 15.10 1 2 101000
   9059   2.891962 40.401409 0.51 0.45 140   80.48 -5937.33 U127500122110 0.55 168 19.00 15.10 1 2 000000
   9037   2.879698 40.398918 0.31 0.31 90 114.11 -5946.28 T2786-00666-1 0.14 266 9.49 7.65 2 9 111111

Here are the J, H, and K images of this area, with this faint source marked by a green circle with a cross inside, the good detection of the nearby bright source marked by the green circle at its center, and all other 'good' sources marked by green circles and artifacts found by MAPCOR marked by red circles.

It looks like this faint source is real.

The MAPCOR confusion radii for the bright star are given below, using the instrumental (uncalibrated, un-normalized) Read1 mags, as MAPCOR does. The distances from the faint source to the bright one in each band (before band-merging) are also indicated; obviously the faint source is outside the present confusion radii and therefore was not marked as confused.

band	instrumental mag	confusion radius (")	distance (")
J	4.824	22.92	34.47
H	4.418	22.90	34.38
K	3.980	27.51	34.85

The J band 'pixel flag' for the faint source indicates that no frame had good SNR magnitude data ("N out of M" was 0/6). The standard aperture mag, however, is 15.583, which is not bad, though it's probably affected by the bright star's halo in its sky background. The H band 'pixel flag' was the same as the J band, 0/6, but the PSF mag was 99.999, which indicates that for some reason PSF fitting didn't work at all in this band. The standard aperture mag, however, is 15.107. The K band 'pixel flag' indicates N/M = 0/5. Also, you can't find it in the data above, but the x and y pixel position errors in both bands, for the J and H source we are concerned about, were each 3.0 pixels, which is very large. The errors are 0.24 pix for the K-only source, which is in the normal regime.

Source 2:

Here is the pertinent J and K band s035.cal file data (it was not detected in H so I removed that info) for this source, followed by the best (Read1) extraction of the bright source very nearby, split into 3 lines each:

| IDNum|    RA    |    Dec   |EMaj|EMin|Ang| X_u |   Y_u   | AssnCatID | R | Th| Mag1| Mag2|O|I|CnfFlg|
   4528   0.827970 38.241749 0.51 0.51 90   -9.62 -13719.90      n/a      0.00   0 99.99 99.99 0 0 010010
   4541   0.832097 38.244034 0.12 0.12 90 -21.29 -13711.68 T2781-00552-1 0.11 122 9.95 8.56 2 8 111111

In the J and K bands it looks like this faint source is real (though there's still a possibility it could be a noise spike in the halo); it's much more difficult to see it in the H band image.

The J and K band MAPCOR confusion radii for the bright star are given below, using the instrumental (uncalibrated, un-normalized) Read1 mags, as MAPCOR does. The distances from the faint source to the bright one in each band (before band-merging) are also indicated; obviously the faint source is outside the present confusion radii and therefore was not marked as confused.

band	instrumental mag	confusion radius (")	distance(")
J	6.438	11.02	14.26
K	5.689	9.70	14.59

The J band 'pixel flag' for the faint source indicates that there was one frame with at least one bad pixel, but the other 5 frames had good (adequate SNR) magnitude data ("N out of M" was 5/5). The K band 'pixel flag', however, shows N/M = 2/6 and the PSF mag was 99.999, which indicates that for some reason PSF fitting didn't work at all in this band. The standard aperture mag, however, is 14.244, which is quite respectable, though it's probably affected by the bright star's halo in its sky background. Also, you can't find it in the data above, but the x and y pixel position errors in both the J and K bands were each 3.0 pixels, which is very large.

Conclusions:

The photometry for both of these sources is almost certainly affected by the nearby bright star, which argues for enlarging the confusion radius as we are planning to do. However, the present confusion flagging algorithm would have marked both of these sources with a flag of "2", because they are Read2-only sources near stars that have other Read2-only sources (mostly artifacts) inside the confusion radius. Remember, at present, only a single Read2 source inside the confusion radius, and all Read1 sources inside the radius, receives a flag of "1" and are allowed into the catalog. Therefore, if we want to worry about completeness around bright stars, I think we need to revisit the flagging algorithm when we determine the new confusion radii.

Perhaps there should be two confusion radii per source, an inner radius and an outer radius, for each parent star. The inner radius, meant to find false sources, would be scaled with the mag of the parent, but might also depend on whether the parent was saturated in Read2 or Read1; any Read2 source inside this radius would receive a flag of "2" and would be considered a false source. Any Read1 source is still likely to be a real source and would therefore receive a flag of "1", as it does now. The outer radius, meant to find real sources with photometry affected by the parent, would also be scaled with parent mag; any source between the inner and outer radii would be considered a real source and receive a flag of "1".

Another option that has been mentioned before is to make the confusion flag value depend on the difference between the parent and child mags. If implemented simplistically, this algorithm would certainly have still marked these sources above as artifacts because of their very faint mags; even if their standard aperture and PSF mags were considered, the mag differences would have been 10-13 mag in the first case, and 8-9 mag in the second.