This Tech Note discusses measurements of magnitudes taken from Tom Droege's Mark IV unit "TOM1" during the period Julian Date 2,452,605 to 2,452,656 (Nov 26, 2002 to Jan 16, 2003). The measurements were processed through the Mark IV pipeline version 0.3. Tom distributed these (and other data) on a set of 5 CD-Roms labelled "TR1" to "TR5". This note concentrates on the contents of the first CD-Rom, "TR1".
The Mark IV unit TOM1 pointed near the celestial equator, covering an area between -5 and +15 degrees Declination. I ignored the area between Right Ascension 300 and 360 degrees, considering only RA between 0 and 160 degrees.
The cameras were set to run all night long, even if clouds arrived. Tom removed data which was obviously bad when he reduced the images into starlists, but it is possible that some periods of cloud and bright sky might remain. How can we remove data taken during those periods?
I wrote a script which uses the .list files generated by the pipeline to check the quality of a night. The script picks out for each image
Here's an example of a good night (2618 = 2,452,618). You can click on the image to download a Postscript version of the graph, which shows more detail. The peak in number of objects is due to the transit of the Milky Way across the Mark IV's scanning area.
And this is a poor night, with clouds coming and going: night 2606.
I examined a plot of this sort for each night of data and selected segments of "good" time: in some cases entire nights, in other cases stretches of a few hours. The "good" time was roughly half of the total time. In all subsequent analysis, I ignored any data taken outside of these good periods of time.
The pipeline derives a photometric solution of the form
calibrated mag = raw mag + a + b * (raw color) + k * airmass
to the frames acquired during each night.
The extinction coefficient k is fixed at some fiducial
value: Tom used kv = 0.20 and ki = 0.06.
Each frame is allowed its own zero point a, but a single
color term b is calculated in each passband for the entire night.
I thought that perhaps nights with a mixture of good and bad measurements might lead to poor values of the color term «b, which would then ruin the photometry of measurements made during the good periods. Therefore, I tried this rather complicated procedure:
However, I didn't see much of a difference in the final values as a result of this extra work. The scatter in measurements of bright stars remained roughly the same size as in the original calibration. It appeared that this scatter was due to errors which depended on a star's location in the field of view, rather than to small variations in the color term from night to night.
So, I gave this idea up.
The TOM1 unit followed a grid-based survey strategy during the period under consideration: it pointed at positions spaced by regular amounts in both RA and Dec, with a small amount of overlap between adjacent fields. The field centers were approximately the same on each night during this period. As a result, a star might fall into one of categories:
I collected measurements of each star together into a single big table, so that I could measure the mean and stdev for each star. I also determined the distance of each star from the center of its image for each measurement. Using a small sample of 4 good nights, I discarded stars which appeared fewer than 3 times, and split the remainder into two groups:
I then plotted the scatter from the mean versus magnitude for stars in these two groups. Look at the graphs in V-band:
And now I-band:
It is clear that the scatter is smaller for the "clustered" samples. This suggests that there is a position-dependent error in the magnitude measurements.
Another clue comes from comparing the TASS magnitudes to those in its parent Tycho2 catalog. Here is a graph showing difference (Tycho2 - TASS) as a function of position on the chip, in V-band:
And in I-band:
It is clear that there is a significant systematic error in the V magnitudes which depends on a star's position on the focal plane. The situation in I-band is less clear: there are differences, but they are at the level of one standard deviation or less for the most part.
I selected a subset of stars to characterize the spatial error in the V-band measurements.
I then compared the mean measured magnitudes of these stars against their values in the Tycho2 catalog. I fit the differences to a simple linear correction term:
error = a + b*x + c*y
where "x" is the location on the focal plane in RA (running
from about -2 degrees to +2 degrees),
and "y" is the location on the focal plane in Dec (also running
from -2 to +2 degrees).
An unweighted least-squares fit yields
V-band a = 0.0103 b = 0.0222 c = 0.0340
I-band a = 0.0026 b = -0.0015 c = 0.0021
The V-band terms are much larger than the I-band terms, as one expects from the graphs of residuals.
I decided to make no corrections to the I-band magnitudes. However, I did apply a correction to the V-band magnitudes in the original .cal files produced by the pipeline. After making the corrections, I again matched up all observations of each star and calculated the mean and stdev from the mean. The corrected data should show a smaller scatter. Look for yourself (and note that the corrected data have been shifted by 0.8 magnitudes on the graph for clarity):
Yes, the corrected magnitudes do show a smaller scatter from the mean. That's good. Here are some numbers, showing the mean difference between Tycho2 and TASS magnitudes as a function of magnitude for the "clustered" subset of stars:
orig V cor V I
mag range N mean sig N mean sig N mean sig
----------------------------------------------------------------------------
8.5 - 9.0 1542 0.016 0.046 1631 0.001 0.031 3380 -0.038 0.092
9.0 - 9.5 4526 0.013 0.055 4580 -0.002 0.037 4088 -0.000 0.086
9.5 -10.0 5401 0.004 0.066 5872 -0.007 0.050 3985 0.031 0.097
-----------------------------------------------------------------------------
Once we have removed systematic errors from the Mark IV data, we can transfer the calibration from Tycho2 to Landolt. Keep in mind that the Tycho2 "I-band" values are really estimates based on an extrapolation from the Bt and Vt passbands.
I matched up the TASS measurements against stars in Landolt's equatorial standards. There were 469 measurements of stars in common. Here are graphs showing the differences (Landolt - TASS) as a function of magnitude:
And the differences as a function of color:
Brief notes:
The differences between the TASS and Landolt magnitudes are a bit smaller than those from the Mark III systems at comparable magnitudes.