TN 0056: Description of TASS tenxcat catalog

Michael Richmond (on behalf of the TASS Collaboration)
July 10, 1999
Revision: #5 990905
Key Words: computation, catalogs, astrometry, photometry

At long last, The Amateur Sky Survey (TASS) is ready to release its first catalog to the community. Let me very briefly explain a bit about TASS, then describe the catalog and its properties. I'll finish by telling you how to access this catalog. But first, the executive summary, so you can decide if it's worth reading any further:

Quick summary:

If you are interested in learning more about the catalog, read on. Here's a Table of Contents:

For more information on the TASS Mark III project, see the TASS home page at http://stupendous.rit.edu/tass/tass.shtml


TASS Mark III cameras

The Amateur Sky Survey (TASS) is a loose confederation of amateur and professional astronomers, organized and run over the Internet. TASS member Tom Droege designed and constructed a series of wide-field electronic cameras. The Mark III unit consists of three CCD cameras in a single mount. Each camera consists of a cheap 135-mm telephoto lens, an astronomical filter, and a Kodak KAF-0400 CCD chip. The CCDs are operated in drift-scan mode: they stare at a fixed position, and clock charge across their arrays at the sidereal rate. Each camera thus creates a long image, about 3 degrees wide from north to south and as long east/west as the night is dark and clear. One camera points due south, one 15 degrees to the east, and the third 15 degrees to the west. Stars drift through the eastern camera, then, an hour later, through the southern camera, and one more hour later, through the west camera. The cameras are aimed near the celestial equator.

Each image has an effective exposure time of roughly 8 minutes. The stellar images are sometimes slightly elliptical, due to a small mismatch between the sidereal and readout rates. The point-spread-function (PSF) may be as sharp as 1.2 pixels full-width at half-maximum, but not always. The camera lenses focus light best in the V-band; in the I-band, some create a wide halo around a sharp core. The image scale is about 14 arcseconds per pixel; any objects closer than about 20 arcseconds appear as a single blob.

Three sites produced the observations included in the July 1999 release of tenxcat: Batavia, IL (Tom Droege), Dayton, OH (Glenn Gombert), Cincinnati, OH (Mike Gutzwiller). All sites are in the suburbs, so the skies are bright. The Dayton and Cincinnati sites used an arrangement of filters I-V-I (east-south-west), whereas the Batavia site used V-R-I. Therefore, tenxcat is dominated by I-band measurements, with a slightly smaller number of V-band and a much smaller number of R-band data.

The September 1999 release of tenxcat contains information based on observations collected between Oct 15, 1996 and Nov 19, 1998. The number of individual detections reported during this interval was

Many of these detections are due to noise, passing aircraft, satellites, etc. The tenxcat catalog is derived from a subset of detections which we deem especially reliable.


Overview of the tenxcat catalog

The tenxcat catalog contains a subsample of objects detected during the Mark III survey period: only objects detected on at least 10 different occasions (where detections by different cameras at one site on a given night are counted as independent). For example,

The September 1999 release of the catalog contains 367,241 stars, of which

Roughly speaking, the catalog contains information on stars with magnitudes between 7 and 14. The figure below shows a histogram of magnitudes in each passband.

Because cameras at different sites were aligned independently, there is a ragged range of Declinations covered by tenxcat, from about -5 to +2 degrees. The coverage in Right Ascension is very uneven (due largely to seasonal weather), with very little coverage between 120 and 200 degrees.


How to access the tenxcat catalog

One way to access the tenxcat catalog is through an HTML form which permits the user to request data in a small area of the sky, either as a list, a chart, or both. You can find the form at

http://stupendous.rit.edu/tass/www_scripts/make_chart.html

There are several other tools one may wish to use in conjunction with tenxcat: look at http://stupendous.rit.edu/tass/dbms_access.html for a brief description and links to HTML forms.

Users with local copies of Postgres may access the TASS archive directly over the Internet:

The TASS archive contains a form of the tenxcat catalog, as a Postgres table called "tenxcat"; information stored in this table has somewhat bizarre units (see a description of the units below).

One may also download the complete tenxcat catalog, with user-friendly units, as a flat ASCII file. The file is about 52 Megabytes in size; read instructions for downloading a gzip'd version (only 9.9 Meg) at

http://stupendous.rit.edu/tass/catalogs/catalogs.html#tenxcat

There is a smaller subset of the entire tenxcat which conists of stars which

Following suggestions from a potential user of the information, this subset has a somewhat different format than the full catalog. Read the description of the high-quality color subset for the full story, but in brief:

This high-quality color subset contains 117,503 stars. The flat ASCII version of the subset is 15 MB in size, and the gzip'd version 2.7 MB. You can download the data from the

http://stupendous.rit.edu/tass/catalogs/catalogs.html#tenxcat


Description of each field in the raw database table

Tenxcat is stored as a table in a Postgres database. The table has a number of fields, described below. Some fields contain cross-identifications of objects in other catalogs; see the note on 'matching' for the details.

Fld# Field name Type Description
1 tass_id int4 Identification number of star in the Mark III database. This number will not change if new stars are added. One can use this ID to request information on a specific star from other tables in the database.
2 ra int4 The Right Ascension of the star, equinox J2000, in units of milliarcsec to the east of RA = 0. Thus, a star at RA = 3.5 degrees would have a value of 12,600,000. This value is an unweighted average of the positions in all detections of the star.
3 dec int4 The Declination of the star, equinox J2000, in units of milliarcsec to the north of Dec = 0. Thus, a star at Dec = -3.5 degrees would have a value of -12,600,000. This value is an unweighted average of the positions in all detections of the star.
4 V int2 The mean V-band magnitude of the star, in millimagnitudes. Thus, a star at with a mag V=8.5 would have a value of 8500. This is the unweighted average of all valid magnitude measurements of the star in the V passband. This field is NULL for a star with no V data.
5 sigV int2 The standard deviation of measurements from the mean V-band magnitude of the star, in millimagnitudes. Thus, a star at with a stdev of 0.03 mag would have a value of 30. This field is NULL for a star with no V data, and is also NULL for a star with 1 V datum.
6 nV int2 The number of measurements of the star in the V-band.
7 R int2 The mean R-band magnitude of the star, in millimagnitudes. Thus, a star at with a mag R=8.5 would have a value of 8500. This is the unweighted average of all valid magnitude measurements of the star in the R passband. This field is NULL for a star with no R data.
8 sigR int2 The standard deviation of measurements from the mean R-band magnitude of the star, in millimagnitudes. Thus, a star at with a stdev of 0.03 mag would have a value of 30. This field is NULL for a star with no R data, and is also NULL for a star with 1 R datum.
9 nR int2 The number of measurements of the star in the R-band.
10 I int2 The mean I-band magnitude of the star, in millimagnitudes. Thus, a star at with a mag I=8.5 would have a value of 8500. This is the unweighted average of all valid magnitude measurements of the star in the I passband. This field is NULL for a star with no I data.
11 sigI int2 The standard deviation of measurements from the mean I-band magnitude of the star, in millimagnitudes. Thus, a star at with a stdev of 0.03 mag would have a value of 30. This field is NULL for a star with no I data, and is also NULL for a star with 1 I datum.
12 nI int2 The number of measurements of the star in the I-band.
13 raw_flag char If '1', this star's magnitude values are not corrected for color terms (usually because it was not observed in both V and I). If a zero ('0'), the star's magnitude has been corrected for color terms.
14 flag2 char If '1', this star was part of a spurious pair (see discussion below), but has been merged with its spurious companion. Detections of both members of the pair have been combined.
15 flag3 char If '1', this star is part of a spurious pair (see discussion below), and it was not possible to merge it with its companion. The other member of the pair must also appear in tenxcat, and its entry will also have flag3 set to '1'.
16 flag4 char Not currently used. Should be NULL.
17 flag5 char Not currently used. Should be NULL.
18 flag6 char Not currently used. Should be NULL.
19 gsc_id char(10) The ID of the matching star in the HST Guide Star Catalog, version 1.1. The ID has two parts: the "small region number" (4 digits), identifying which of roughly 9500 sections of the sky contains the star, and the "number-within-region" section (5 digits), The two pieces are always separated by a hyphen. If there is no matching GSC star, the field is NULL.
20 ac2000_id int4 The ID of the matching star in the Astrographic Catalogue 2000, which is being constructed by the US Naval Observatory. Actually, this ID was taken from the ACT catalog, a prequel to the AC2000. If there is no matching ACT star, the field is NULL.
21 hd_id int4 The ID of the matching star in the Henry Draper Catalogue. Actually, this ID was taken from the ACT catalog. If there is no matching HD star, the field is NULL.
22 spectype char(4) The spectral type of the star, from the Henry Draper Catalogue. If there is no matching HD star, this field is NULL.
23 iras_id char(12) The name of a matching source in the IRAS Point Source Catalogue. If there is no matching IRAS source, this field is NULL.

One may request data from tenxcat in this format, but it is unlikely that many people will wish to do so. Most people will probably request the "user-friendly format", described below.


Description of each field in the user-friendly output

Fld# Field name Type Description
1 name int
I8
or
string
A15
Depending on the user's preference, this may be either the tass_id value (described above), which looks like this:
234321

or the ID of the star in the GSC version 1.1:
0323-00222

or, if there is no matching GSC star, a name in the spirit of the proper IAU designation which looks like this:
HHMMSS.s+DDMMSS
In order to make this last name into a real IAU designation, one must add the prefix "TASS J"; for example,
TASS J023553.2-020325
2 ra float
F8.4
The Right Ascension in decimal degrees, equinox J2000. This value is an unweighted average of the positions in all detections of the star.
3 dec float
F8.4
The Declination in decimal degrees, equinox J2000. This value is an unweighted average of the positions in all detections of the star.
4 vmag float
F6.3
The V-band magnitude. This value is an unweighted average of the measurements in all detections of the star. A blank value means no data.
5 vsig float
F6.3
The standard deviation from the mean V-band magnitude. A blank value means no data (or only a single V measurement).
6 nv int
I3
The number of detections in V-band.
7 rmag float
F6.3
The R-band magnitude. This value is an unweighted average of the measurements in all detections of the star. A value of 30.000 means no data.
8 rsig float
F6.3
The standard deviation from the mean R-band magnitude. A blank value means no data (or only a single R measurement).
9 nr int
I3
The number of detections in R-band.
10 imag float
F6.3
The I-band magnitude. This value is an unweighted average of the measurements in all detections of the star. A blank value means no data.
11 isig float
F6.3
The standard deviation from the mean I-band magnitude. A blank value means no data (or a single I-band measurement).
12 ni int
I3
The number of detections in I-band.
13 flags 6 ints
6 * I1
This field contains six values, each a single digit, either "1" or "0". See a detailed explanation below.
14 ac2000_id int
I7
The ID of the matching star in the Astrographic Catalogue 2000, which is being constructed by the US Naval Observatory. Actually, this ID was taken from the ACT catalog, a prequel to the AC2000. If there is no matching ACT star, the field is NULL.
15 hd_id int
I6
The ID of the matching star in the Henry Draper Catalogue. Actually, this ID was taken from the ACT catalog. If there is no matching HD star, the field is NULL.
16 spectype string
A4
The spectral type of the star, from the Henry Draper Catalogue. If there is no matching HD star, this field is NULL.
17 iras_id string
A12
The name of a matching source in the IRAS Point Source Catalogue. If there is no matching IRAS source, this field is NULL.

Here is a detailed description of the flags encoded in field 13 of the user-friendly format. The field contains six single-digit values, in which each digit is either '0' or '1'.

Some examples:


Description of each field in the high-quality color subset

Fld# Field name Type Description
1 name string
A15
This is usually the ID of the star in the GSC version 1.1:
0323-00222

but, if there is no matching GSC star, a name in the spirit of the proper IAU designation which looks like this:
HHMMSS.s+DDMMSS
In order to make this last name into a real IAU designation, one must add the prefix "TASS J"; for example,
TASS J023553.2-020325
2 ra string
A10
The Right Ascension in sexigesimal notation:
HH:MM:SS.s
3 dec string
A9
The Declination in sexigesimal notation:
+DD:MM:SS
4 vmag float
F5.2
The V-band magnitude. This value is an unweighted average of the measurements in all detections of the star.
5 vsig float
F5.2
The standard deviation from the mean V-band magnitude.
6 nv int
I3
The number of detections in V-band.
7 V-R color float
F5.2
The (V-R) color. A blank value means no data in either V or R, or both.
8 V-R sigma float
F5.2
The standard deviation of the (V-R) color, based on the uncertainties in the V and R magnitudes. A blank value means no data. When only a single R-band measurement exists, it is assumed to contribute exactly 0.10 mag to the uncertainty in the (V-R) color.
9 nr int
I3
The number of detections in R-band.
10 V-I color float
F5.2
The (V-I) color of the star.
11 V-I sigma float
F5.2
The standard deviation from the mean (V-I) color, based on the uncertainties in the V and I measurements.
12 ni int
I3
The number of detections in I-band.
13 flags 3 ints
3 * I1
This field contains three values, each a single digit, either "1" or "0"; they denote the "raw_flag", "flag2" and "flag3" values. See a detailed explanation below.
14 ac2000_id int
I7
The ID of the matching star in the Astrographic Catalogue 2000, which is being constructed by the US Naval Observatory. Actually, this ID was taken from the ACT catalog, a prequel to the AC2000. If there is no matching ACT star, the field is NULL.
15 hd_id int
I6
The ID of the matching star in the Henry Draper Catalogue. Actually, this ID was taken from the ACT catalog. If there is no matching HD star, the field is NULL.
16 spectype string
A4
The spectral type of the star, from the Henry Draper Catalogue. If there is no matching HD star, this field is NULL.
17 iras_id string
A12
The name of a matching source in the IRAS Point Source Catalogue. If there is no matching IRAS source, this field is NULL.


How are 'matches' between TASS and other catalogs defined?

In order to find "matches" between TASS detections and entries in other catalogs, we used a simple algorithm:

  1. set a minimum matching distance for the other catalog
  2. for each TASS detection, look through the other catalog for objects closer than this minimum

The TASS cameras have pixels about 14 arcseconds on a side, and measure positions for isolated stars to a few arcseconds. We adopted minimum matching distances of

We made no check against several TASS detections matching the same source from another catalog.


What are "spurious pairs"?

The pixel size of the Mark III images is about 14 arcsec. When we created the database, we needed to identify detections of the same star in different images. We used the following rule:

       If a new detection falls within 15 arcsec of an existing
       entry in the database, combine it with the existing entry.
       Otherwise, make a new entry for this detection.

Occasionally, some image would show a star at a position more than 15 seconds from its current mean position in the database. In such cases, we created a new entry in the database for the star -- a mistake. We discovered this error when charts based on an early version of tenxcat showed a very large number of close double stars with very similar magnitudes. We call these cases in which one or more detections of a star failed to match up with an existing entry spurious pairs.

Actually, there were a number of clumps containing more than two "close" objects, where "close" means "within 15 arcseconds of another object in the catalog." Very roughly, we found

Inspection of a few cases revealed that some of the clumps were due to truly extended objects: galaxies and clusters. We decided to ignore all cases of more than 2 objects.

We devised the following algorithm for finding truly spurious pairs; it should leave true pairs untouched.

     Make a preliminary version of the 'tenxcat' catalog

     For each star in the preliminary catalog
        is there another star within 15 arcsec?
        if no, skip to next star
        if yes,
          do the two stars ever appear in the same image?
          if yes, skip to next star
          if no,
            do the two stars have magnitudes which differ by < 0.5 mag?
            if no, skip to next star
            if yes, mark as a spurious pair

Using this algorithm, we identified about 15,000 spurious pairs in the preliminary version of tenxcat. We attempted to merge them together in the following manner:

We were unable to merge the two members in about 20 percent of the cases: usually because one member had both V and I magnitudes -- allowing us to correct its values for color terms -- but the other did not.

If we were able to merge the two members, we

If we were unable to merge the two members, we

The result is an improved version of tenxcat, which still contains some spurious pairs ... but not as many as the preliminary version. We distribute only the improved version.


How were the images reduced into lists of stars?

Mike Gutzwiller wrote the software we use to correct the raw images, find stars, measure their properties, perform astrometric calibration and an initial photometric calibration. Look at Mike Gutzwiller's TASS Software page for a brief description of each program, source code, and executables.

The steps involved in this reduction are

  1. create a dark vector from a dark image
  2. subtract the dark vector from the data images
  3. create a flatfield vector from a dark-subtracted image
  4. correct each image by subtracting the dark vector and dividing by the (normalized) flatfield vector
  5. detect stars in the corrected image; we do so by creating a 2-D gaussian PSF for the frame, based on very bright stars, then convolving the frame with the PSF and looking for peaks
  6. measure the position of each star in pixel coordinates (row, col) within the image
  7. measure the instrumental magnitude of each star, using an elliptical aperture matched to the size of the PSF
  8. match detected stars to the Tycho catalog and determine the astrometric transformation which converts (row, col) to (RA, Dec)
  9. calculate the (RA, Dec) of each detected star
  10. perform a preliminary photometric calibration, as follows:

The output of this entire process is a list of stars detected in each image. The list contains (RA, Dec) positions for each star, plus a preliminary magnitude on the Johnson-Cousins system.

The TASS members at each site with a camera ran all their images through this procedure locally. They then sent the starlists to RIT, where the data was entered into the master TASS database. Chris Albertson wrote the software which read the star lists and added their information to the database. In short, the starlist from each image was compared to the master TASS catalog: stars within 15 arcseconds of existing entries in the catalog were add as instances of those entries, while isolated stars gave rise to new entries in the master catalog.


Correcting for color terms in individual cameras

After all the data had been merged together in the master database, it was possible to examine it for small systematic errors. One source of error is a mismatch between the combined overall spectral response of a telescope, filter and CCD, and the standard astronomical passband. Small mismatches cause stars of different colors to appear brighter or fainter than they ought. The usual practice is to look for residuals as a function of color, then fit those residuals with a simple linear function of color.

Recall that there were 9 cameras involved in the Mark III survey:

We compared TASS measurements of stars against Landolt's equatorial catalogs of UBVRI standard stars. The V-band measurements showed no significant offset in the mean, nor any significant color dependence. We therefore made no correction to the V-band measurements. In the R and I bands, however, there were both small offsets in the mean (always in the sense that TASS measurements were slightly fainter than Landolt measurements), and small trends in residuals as a function of (V-I) color. See TASS Technical Note 54 for details, graphs, and tables.

We determined the following color terms, based on stars brighter than mag 11 and using a color which was the mean TASS V value (based on measurements from all cameras) minus the mean TASS I value (based on measurements from all cameras). In the equations below, uppercase values represent corrected single magnitude measurements, and lowercase values represent raw single magnitude measurements.

                         R-band 

  camera H1 (Batavia):  R = r + 0.01209 - 0.073704*(color)


                         I-band

  camera B0 (Dayton):   I = i - 0.0563  - 0.018676*(color)
  camera B2 (Dayton):   I = i - 0.0683  + 0.040971*(color)

  camera D0 (Cinn):     I = i - 0.0844  + 0.06116 *(color)
  camera D2 (Cinn):     I = i - 0.1001  + 0.08510 *(color)

  camera H2 (Batavia):  I = i + 0.0121  - 0.073704*(color)
  

After making the corrections to each individual measurement, we then re-calculated the mean magnitude of each star in all passbands. In cases of spurious pairs, we made the corrections to each member of the pair separately, based on its own color, re-calculated mean magnitudes for each member separately, and finally determined a weighted mean magnitude for the merged entity.

The corrected tenxcat magnitudes agree well with Landolt magnitudes, with little dependence on color:


How good is the astrometry?

We initially calibrated the positions of each detection against the Tycho catalog. As we merge successive observations of a source, we calculate a running mean for its position. The positions of stars in tenxcat are pretty good:

              50%    match Tycho positions to   0.80  arcsec
              67%                               1.16
              90%                               2.41

Watch out! Remember that the pixel size in the initial images is about 14 arcsec. Close pairs or groups of stars are liable to be detected as a single source with wierd properties, or ignored altogether. We have discovered another shortcoming in the catalog: during the merging process, we assumed that any detection within 15 arcsec of another was the same source. Occasionally, one measurement of a star would differ from the mean position by more than 15 arcsec, and we would create a new entry in the catalog for it. I have attempted to remove some of these spurious multiple stars, but several thousand remain.


How good is the photometry?

We used the Tycho catalog for initial photometric calibration. Note that Tycho contains measurements in a V-like passband and a B-like passband ... whereas our cameras measure V, R, and I. We used an empirical relationship between the Tycho B and V magnitudes to estimate R and I. After comparing the resulting magnitudes to values from Landolt for equatorial standards, we found small color-dependent residuals in the R-band and I-band measurements. We corrected measurements for these effects before placing them into the tenxcat catalog. Comparing our magnitudes to Landolt's reveals the following scatter:

                        TASS vs. Landolt

       passband        bright (mag < 11)      faint (mag > 11)
      ---------------------------------------------------------
          V              0.03 mag                 0.09 mag
          R              0.08 mag                 0.10 mag
          I              0.09 mag                 0.11 mag

Here is a graph showing the difference between Landolt and tenxcat magnitudes on a star-by-star basis. One can see that as one approaches the plate limit (around V = 13.5 and I = 12.6), the errors go way up. The V-band magnitudes for V > 13.5 appear brighter than they ought to be.

Another way to look at the accuracy of the TASS magnitudes is to plot the distribution of differences (Landolt - TASS):

A measure of photometric precision (but not accuracy) is the internal scatter of measurements from their mean. We calculated the standard deviation from the mean magnitude for each star in each passband, then found the median value in each passband.

   Median of internal standard deviation

        Mag      V      R      I
      ------------------------------
        6.00   0.033  0.000  0.046 
        6.50   0.039  0.041  0.049 
        7.00   0.044  0.033  0.046 
        7.50   0.041  0.032  0.044 
        8.00   0.039  0.034  0.047 
        8.50   0.037  0.034  0.050 
        9.00   0.040  0.033  0.055 
        9.50   0.042  0.035  0.065 
       10.00   0.046  0.037  0.080 
       10.50   0.054  0.041  0.102 
       11.00   0.067  0.050  0.136 
       11.50   0.090  0.068  0.177 
       12.00   0.125  0.095  0.218 
       12.50   0.175  0.140  0.257 
       13.00   0.228  0.193  0.292 
       13.50   0.275  0.194  0.268 
       14.00   0.302  0.106  0.116 
       14.50   0.173  0.000  0.000 

The table and graph below show that measurements in the I-band contain much larger internal errors than those in V-band and R-band. We suspect the main contributions to this increased scatter are: