The service described below has been discontinued. Sorry. Michael Richmond, Sept 7, 1997.
I have written some routines to perform two kinds of astrometric calibration of starlists, and made it possible for people to use the software by E-mailing their data to my machine. Let me explain a few terms before I get to the good stuff.
While I have verified that this procedure succeeds most of the time on some recent TASS images I took and reduced myself, I have discovered that it does not succeed on some starlists which Tom Droege acquired in June, 1996, and distributed. Therefore, I can make no guarantees that the code will succeed with your data. Treat this as an ALPHA version of a product.
First, a starlist is an ASCII text file, which contains data in a simple multi-column format. Any line starting with the pound-sign character # is ignored, as is any completely blank line. All other lines must contain the same number of columns of information, which may be numbers or letters. Here's an example of a simple starlist that might be generated by some source-extraction program:
#
# ID row col sky skysig mag magerr
#
1 3.83 231.21 206 20.59 17.133 0.080
2 3.63 667.10 202 14.85 18.143 0.185
3 9.63 540.46 202 116.79 16.725 0.057
4 13.52 273.85 197 254.42 17.433 0.101
5 16.22 96.82 202 24.82 16.420 0.044
6 16.44 458.32 194 183.74 16.592 0.050
In order to use this software, you should restrict such lists
to extend no more than 5 degrees in Right Ascension,
and, in fact, a length of only 3 degrees would be better.
For a TASS-like pixel scale of 13.8 arcsec/pixel, this means
the list should extend over 700-1200 rows;
in other words, only 10-18 minutes of scanning.
Next, let me explain the limitations on this astrometric calibration. I have taken an old version of the Hubble Space Telescope Guide Star Catalog, removed all stars which fall more than five degrees from the celestial equator (i.e. Dec > 5 or Dec < -5) and compressed it. Therefore, my software can only recognize stars which occur within this narrow band. Moreover, in the interests of time and space, I have created a further subset, dividing the sky into 72 chunks of 5 degrees in Right Ascension, and keeping only the brightest 40 stars in each chunk. In order for the field-recognition software to work, there must be substantial overlap between this set of the brightest 40 stars in GSC and the brightest 40 stars in the starlist. Fortunately, for the TASS images I have examined so far, the overlap contains a good number of stars. But data from other telescopes, or data taken under some conditions, may fail to yield enough stars for matching.
My first goal is to figure out where in the sky the telescope was pointing when it observed the objects in the starlist. Here's how: the 40 brightest stars are chosen from the starlist, and a set of all possible triangles are made from this bright, observed subset. I have pre-computed and stored on disk similar subsets of the 40 brightest stars and triangles from each 5-degree chunk along the celestial equator (well, technically, centered on Dec = +00:40:00). The code checks for correspondences between triangles in the starlist and each 5-degree chunk.
If it finds no good match, it quits, and the astrometric calibration has failed.
If at least one good match is found, the code picks the chunk with the best match, and calculates the approximate position of the center of the starlist within the chunk. The center of the starlist is simple the average of the extreme rows, and the average of the extreme columns. It then generates a new, much larger set of stars from the Guide Star Catalog, around the new central position. This larger set of stars is compared to the starlist, and again the code calculates a new central postion. In addition, it calculates a TRANS structure which converts observed (row, col) coordinates to (RA, Dec) on the sky, via equations like:
RA = A + B*row + C*col
Dec = D + E*row + F*col
The TRANS conversion
is equivalent to performing translation, rotation,
and scale change, but does not allow for
any distortions across the field of view, or along
the scan direction.
If the first step succeeds, then the code moves on to convert the (row, col) coords of each detected star in the starlist to their (RA, Dec) values. Since the reference catalog is the GSC, the equinox is J2000, but the epoch is around 1975 (I think). This should not be an issue for TASS images.
Given some input starlist, and the location of the columns containing the original (row, col) positions, the code simply reproduces the entire input starlist, replacing the columns
row -> Right Ascension (decimal degrees)
col -> Declination (decimal degrees)
All other information is left as-is.
However, the spacing of the replaced columns may
be slightly changed.
Okay, so how do you actually do this? It involves three steps. Pay attention carefully, as details of the header format will matter!
Step 1: Create a starlist using your favorite astronomical software to extract sources. Make sure that any comments occur in lines which start with the pound-sign character #. Place the following special header at the start of your file, before any data lines appear:
# start # xcol=2 # ycol=3 # magcol=6 # mailaddr=put_you_Email_address_heremust appear in the very first position in each line, and that there must be a single space between the pound sign and the keyword. Note also that there must be no spaces between the keywords, the equals signs, and the values (i.e. between 'xcol', '=', '2'). The lines mean
There is also one optional keyword:
Step 2: Mail the starlist, with header, to
richmond@astro.princeton.edu
with a 'Subject:' line that looks exactly like this:
Subject: tass-astrometry
Now, wait for a reply. If the network is lightly loaded, and my machine isn't busy, you may get a reply within 10 minutes.
Step 3: Examine the returned E-mail message. The 'Subject:' line should indicate if the calibration succeeded or failed.
If the 'Subject:' line looks like this:
Subject: astrometric calibration succeeded
then a correct solution may have been found. You should
look carefully at the first few lines of the mail message
itself. They should look something like this:
# # field center is (J2000) RA=357.52439 Dec=1.00456 # coords calculated on Mon Oct 14 14:44:59 EDT 1996Check that the RA values lies between 0 and 360 degrees, and also that the Dec value is somewhere between -4 and +4 degrees. A poor match can lead the Dec value seriously astray -- if you see a value like "Dec=-12.23343", then you know the matching process failed. You should verify that only the (row, col) coords of your original file have been modified, and that all the other columns contain the same information as the original.
If the 'Subject:' line looks like this:
Subject: astrometric calibration failed
then a correct solution was not found.
You should check carefully to make sure that the header lines
in your original mail message describe the format of the
data correctly.
You may ask, "Why can't I get the code to do this calibration myself?" After all, there are several other software packages I've written which are available. There are two reasons I am not currently distributing the source code for this particular software:
If you really, really want a copy of this yourself, you can try to contact me directly.
Here are the first few lines of a good input file, one that could be mailed to the astrometry server:
# start
# xcol=2
# ycol=3
# magcol=6
# mailaddr=richmond@astro.princeton.edu
#
1 3.83 231.21 206 20.59 17.133 0.080
2 3.63 667.10 202 14.85 18.143 0.185
3 9.63 540.46 202 116.79 16.725 0.057
4 13.52 273.85 197 254.42 17.433 0.101
5 16.22 96.82 202 24.82 16.420 0.044
6 16.44 458.32 194 183.74 16.592 0.050
And here are the first few lines of the result of the successful calibration of the above starlist:
# # field center is (J2000) RA=357.52439 Dec=1.00456 # coords calculated on Mon Oct 14 14:44:59 EDT 1996 1 355.43299 1.58080 206 20.59 17.133 0.080 2 355.43429 -0.09110 202 14.85 18.143 0.185 3 355.45672 0.39459 202 116.79 16.725 0.057 4 355.47040 1.41730 197 254.42 17.433 0.101 5 355.47994 2.09628 202 24.82 16.420 0.044 6 355.48248 0.70968 194 183.74 16.592 0.050Note that the spacing of the "extra" columns in the file has been changed. Note also that there is no identication of the columns -- it is up to the user to remember the meaning of each column in the original starlist.