I. Skipped Single-Band Parents

1. Description

Since DB_MAPCOR parents are very bright, the real ``parent'' sources should be detected in all three bands. However, the pipeline has trouble cleanly detecting and band-merging each component of close visual multiple stars, sometimes resulting in single-band detections of one or more component stars and occasionally a set of false ``detections'' between the components.

DB_MAPCOR interpolates a better magnitude when a parent source was not detected in one band, as follows:
j_m = 2.695 * h_m - 1.695 * k_m
h_m = 0.370 * j_m - 0.630 * k_m
k_m = 1.598 * h_m - 0.587 * j_m

However, when a parent source was not detected in two bands, it does not attempt to use the parent to find artifacts. There are 12 of these single-band parents that have no other ``good'' parent duplicate source in the database; all but one are members of close visual multiples. (``Good'' parents are found in the set of released scans, not already marked as artifacts, and have detections in two or three bands.) These parents and their unmarked artifacts are listed in section 2 below with comments and images.

MAPCOR used these sources as ``parents'' in one band during pipeline processing, so some of their artifacts have been flagged in that band; but since DB_MAPCOR did not used them as ``parents'', some diffraction spike artifacts found in nearby scans were not flagged. Only 28 artifacts from these parents are found in the PSC, for two reasons:

• None of these parents are very close to scan edges, so they did not create any confusion artifacts in nearby scans.
• For each of the 11 skipped parents that are members of close visual multiples, there was at least one two- or three-band extraction of another component of the group or in the middle of the components. Thus, MAPCOR and DB_MAPCOR found some artifacts that happened to fall in the area masked by those two- or three-band parents.

2. Data

In the images below, the positions of the skipped parents are marked with a green star, and the positions of the missed diffraction spike artifacts are marked with a red ``x''. A source described as a ``good extraction'' is found in the set of released scans, is not already marked as an artifact, and has detections in two or three bands. The Src column indicates whether the source is a skipped parent, ``par'', or a missed artifact, ``art''. The missed artifacts include both the current cc_flg values and the values that they would have received if the parents had not been skipped, cc_true; the parents have blanks in cc_true.
 

Src	cntr	hemis	date	scan	ra	dec	j_m	h_m	k_m	rd_flg	bl_flg	cc_flg	cc_true	ndet
par	103241977	s	981010	54	347.998278	-11.933574	7.061	6.646	4.917	001	001	00c		000067
par	*2278612	s	980624	147	347.998348	-11.933611	7.054	6.756	4.887	001	001	00d		000044
art	2278639	s	980624	147	347.998276	-11.928149	13.151	14.580	12.683	020	020	0c0	0D0	005500
art	103241953	s	981010	54	347.998208	-11.924371	14.793	15.614	14.283	020	010	0c0	0D0	001500
art	103242016	s	981010	54	347.998168	-11.939388	14.373	13.170	12.657	200	300	c00	D00	660000
art	103242025	s	981010	54	347.998140	-11.942783	14.729	15.263	14.188	020	020	0c0	0D0	004500
Comments: One component of multiple; the second component has two good extractions, cntrs 2278598 & 103241991. *Source 2278612 is found in the catalog generation database but was removed from the catalog during duplicate source resolution.

 

Src	cntr	hemis	date	scan	ra	dec	j_m	h_m	k_m	rd_flg	bl_flg	cc_flg	cc_true	ndet
par	217158358	s	990112	26	97.206584	-7.034617	6.850	6.832	4.769	001	001	000		000066
art	217158303	s	990112	26	97.206530	-7.043774	13.756	15.296	13.868	020	010	0c0	0D0	004500
art	217158304	s	990112	26	97.204465	-7.043096	13.404	13.241	14.884	002	001	00c	00D	000005
art	217158307	s	990112	26	97.204403	-7.042271	13.179	14.639	13.254	020	010	0c0	0D0	005500
art	217158310	s	990112	26	97.204456	-7.040927	12.690	12.972	14.338	002	001	00c	00D	000055
art	217158398	s	990112	26	97.196955	-7.032991	12.754	13.109	14.644	002	002	00c	00D	000046
art	217158399	s	990112	26	97.195936	-7.032983	12.956	13.276	14.579	002	002	00c	00D	000005
art	217158402	s	990112	26	97.198410	-7.032949	12.427	12.774	14.704	002	001	00c	00D	000056
art	217158446	s	990112	26	97.204538	-7.025129	12.954	14.824	13.232	020	010	0c0	0D0	005600
art	217158456	s	990112	26	97.204603	-7.022910	13.713	15.116	13.689	020	010	0c0	0D0	001600
Comments: One component of multiple; the second component, cntrs 217158403 and 217158385, and third component, 217158367, 217158372, and 217158373, are all flagged as confusion artifacts. The only good extraction 217158378 is in the middle of the group (mags 3.72, 3.52, 4.08).

 

Src

cntr

hemis

date

scan

ra

dec

j_m

h_m

k_m

rd_flg

bl_flg

cc_flg

cc_true

ndet

par

289383079

s

990227

80

190.316455

-13.013838

6.568

6.930

4.722

001

001

00c

000046

Comments: One component of multiple; the second component has a good extraction, cntr 289383099.

 

Src	cntr	hemis	date	scan	ra	dec	j_m	h_m	k_m	rd_flg	bl_flg	cc_flg	cc_true	ndet
par	341145237	n	981001	99	28.382560	19.293926	6.933	6.710	4.716	001	001	00d		000046
par	341145270	n	981001	99	28.382574	19.296000	6.917	6.726	4.680	001	001	00d		000046
art	341145233	n	981001	99	28.392755	19.293825	13.788	14.935	13.406	020	010	0c0	0D0	003300
art	341145234	n	981001	99	28.390930	19.293865	13.187	14.454	13.115	020	010	0c0	0D0	004400
art	341145275	n	981001	99	28.398193	19.296101	16.920	14.993	14.467	200	100	c00	D00	160000
art	341145276	n	981001	99	28.372165	19.296106	13.951	14.958	13.542	020	010	0c0	0D0	005500
Comments: Two components of multiple; the only good extraction 341145260 is in the middle of the group.

 

Src	cntr	hemis	date	scan	ra	dec	j_m	h_m	k_m	rd_flg	bl_flg	cc_flg	cc_true	ndet
par	385184050	n	981023	54	110.029590	21.981293	13.251	2.704	6.265	030	010	000		000300
art	385164043	n	981023	53	109.968617	21.981981	16.644	16.166	15.748	222	111	000	DDD	160506
art	385184101	n	981023	54	109.989774	21.983397	16.569	15.384	14.890	200	100	c00	D00	260000
art	385184102	n	981023	54	109.991200	21.983400	16.468	15.888	14.399	200	100	c00	D00	260000
Comments: One component of spectroscopic binary; the second component 385184073 has a good extraction. The 3-band missed artifact 385164043 could be a real source, but it is too faint to be marked as real by DB_MAPCOR.

 

Src

cntr

hemis

date

scan

ra

dec

j_m

h_m

k_m

rd_flg

bl_flg

cc_flg

cc_true

ndet

par

503601069

n

990104

114

102.051265

55.704258

7.085

6.797

4.782

001

001

00d

000066

Comments: One component of multiple; the second component has a good extraction 503601055.

 

Src

cntr

hemis

date

scan

ra

dec

j_m

h_m

k_m

rd_flg

bl_flg

cc_flg

cc_true

ndet

par

599766715

s

990618

8

245.597605

-52.750397

6.963

6.769

2.950

003

001

00d

000004

Comments: One component of multiple; the second component has good extractions 599766693 & 599800179.

 

Src	cntr	hemis	date	scan	ra	dec	j_m	h_m	k_m	rd_flg	bl_flg	cc_flg	cc_true	ndet
par	615163206	n	990403	39	200.984544	54.921768	7.331	6.841	2.821	003	001	000		000004
art	979479789	n	000216	87	201.006294	54.921829	14.126	13.468	14.342	002	001	00c	00D	000055
art	979479792	n	000216	87	201.014101	54.921680	15.724	15.201	15.146	062	002	00c	00D	000016
Comments: One component of multiple; the second component has a good extraction 615163166.

 

Src	cntr	hemis	date	scan	ra	dec	j_m	h_m	k_m	rd_flg	bl_flg	cc_flg	cc_true	ndet
par	1054321630	n	000411	74	207.266744	26.979107	7.084	6.808	4.415	001	001	00d		000056
art	1054321643	n	000411	74	207.275803	26.978880	14.121	15.425	13.534	026	030	0c0	0D0	006600
Comments: One component of multiple; the second component has a good extraction 1054321618.

 

Src	cntr	hemis	date	scan	ra	dec	j_m	h_m	k_m	rd_flg	bl_flg	cc_flg	cc_true	ndet
par	1099436925	n	000506	25	270.376684	21.595833	6.729	6.697	2.540	003	001	00d		000001
art	1099437044	n	000506	25	270.376600	21.580906	16.403	14.236	13.938	200	100	c00	D00	260000
Comments: One component of multiple; the second component also has no good extraction. The only good extraction 1099436940 is in the middle of the group (mags 2.73, 2.53, 2.30).

 

Src	cntr	hemis	date	scan	ra	dec	j_m	h_m	k_m	rd_flg	bl_flg	cc_flg	cc_true	ndet
par	1321822497	n	980927	128	76.142250	15.404170	6.693	6.564	4.738	001	001	000		000066
par	*859750511	n	991218	43	76.142295	15.404105	4.776	4.756	4.723	111	111	000		062636
art	419174136	n	981117	112	76.150385	15.404145	13.565	13.494	14.259	662	002	00c	00D	000044
art	419174137	n	981117	112	76.151647	15.404104	14.122	13.740	14.433	662	002	00c	00D	000055
art	419174140	n	981117	112	76.147564	15.404116	12.195	12.599	13.581	002	002	00c	00D	000055
art	419174143	n	981117	112	76.156197	15.404009	15.554	15.159	15.009	002	001	000	00D	000036
Comments: This is the only skipped parent that is not a component of a visual multiple. The J and H band detections were bandmerged together into source 1321822479, but each band had N/M = 1/1 and thus did not meet the catalog generation database criteria. The duplicate *859750511 is a good source, but its scan is not in the release scan set.

From Davy Kirkpatrick's Sloane correlation analysis, I've identified a bug in v3 MAPCOR persistence searches. MAPCOR read in Read1 parents in the "low-coverage area", the start and end of the scan before and after the 6-frame-deep "full coverage" area. It used these sources as possible parents to find artifacts that could extend into the full-coverage area. (It then discarded these low-coverage sources and did not pass them on to the rest of the pipeline.)

However, there was a bug in the search for Read2-Read1 persistence sources when the parent was in the low-coverage area at the beginning of a south-going scan, and the predicted position of its first persistence source was north of (i.e. before) the first Read2-Read1 source in the scan. (MAPCOR used the minimum and maximum y coordinates of the Read2-Read1 sources to define the Read2-Read1 "full coverage" area, so its full coverage area was almost always smaller than the actual 6-frame sky coverage area.) When the predicted position was north of MAPCOR's "full coverage" area, the initial persistence search would recognize this and set a flag that would prevent MAPCOR from doing further Read2-Read1 persistence searches at later -- in this case, more southern -- frame positions. My intent with this flag was to stop searching when moving away from the full-coverage area, but unfortunately in this case it actually is moving into the full-coverage area.

This bug does not affect persistence searches with parents in, or less than 1 frame away from, MAPCOR's Read2-Read1 full-coverage area, or in north-going scans. It also does not affect searches for Read1 persistence sources, nor any other type of artifact searches (e.g., glints, diffraction spikes, etc.).

2. Data

As far as I know, only one of these sources has been found so far during v3 analysis. Because the data on the number of frames searched for persistence has not been kept in the database, I can't find an easy way to count the number of sources missing persistence probabilities because of this bug. Therefore, at Gene Kopan's suggestion, I counted the number of sources in the working database and the catalog with persistence parents in the low-coverage area to get an idea of the scope of the problem. In the table below, pprob is the persistence probability, WDB is the working database table, and PSC is the point source catalog.

Approximately half of all survey scans, the north-going scans, were not affected by this bug. A more precise count gives us 35309 north-going scans and 35403 south-going scans in the WDB, and 29830 north-going scans and 29901 south-going scans in the PSC scan set. In both cases, the fractions of north- and south-going scans are each 0.50, to within 0.001. If we assume that the north- and south-going scans are distributed randomly over the sky, half of the WDB and PSC sources should be found in north-going scans and the other half in south-going scans. This means that approximately half of this type of persistence sources were flagged correctly, and half were not.

To guess how many unreliable sources, with true pprob > 0.5, and persistence-affected sources, with true pprob <= 0.5, were unflagged, we could get the upper limits from the sources that do have pprob values in the PSC and WDB and have persistence parents from the low-coverage area (see the first two lines of the table below). Only 20-25% of these pprob <= 0.5 sources in the WDB are also found in the PSC, probably because most of them are real sources that happen to fall on a persistence position and thus have duplicate extractions in neighboring scans. These duplicates are likely to be farther from the scan edges and therefore would be selected in the duplicate-resolution processing, and they will either have full-coverage persistence parents if their scans run in the same direction as the scans with low-coverage parents, or will not fall on a persistence position if their scans run in the opposite direction.

However, not all parents will be able to create more than one persistence source, as required for this bug to occur. The numbers of low-coverage parents creating more than one persistence source in the DB are found in the third and fourth lines of the table, and the numbers of persistence sources created or affected by these parents are found in the fifth and sixth lines. Unfortunately, we can only get these counts one band at a time, so the "any band" category is left blank. Since approximately 76% of the sources in the PSC are 3-band sources, these counts will certainly overestimate the number of persistence sources from low-coverage parents.

Assuming that approximately half of this type of persistence sources were flagged correctly, and half were not, and ignoring the fact that most of the sources are 3-band, not single-band, the last two lines of the table are a reasonable estimate of the maximum number of missed persistence sources. A maximum of 94364 WDB sources should have had pprob > 0.5; if about 17%, 24%, and 19% of them in the J, H, and K_s bands respectively are in the PSC, the catalog contains approximately 5621, 9965, and 3758 missed persistence artifacts in J, H, and K_s, or 19344 total. This is 0.0041% of the total number of sources in the PSC (470992970). The maximum number of missed WDB sources that should have had pprob <= 0.5 is approximately 117092; the PSC contains approximately 4130, 10492, and 9273 of these sources in J, H, and K_s, or 23895 total. This is 0.0051% of the total sources in the PSC. Because of the assumptions made here, these numbers are definitely overestimates by at most a factor of 3.