The single frame source extraction process utilizes psfs to estimate the x and y pixel coordinates of local maxima in each flattened frame. These positions of stars in the single frames are used to determine the relative positions of the frames and to position the scans on the sky. The single frame positions and magnitudes for read1 sources are combined and reported in the catalog, while the parameters are re-estimated by PROPHOT for read2 sources.
The position estimation method used in FREXAS for 2MAPPS Version II does not work well for saturated sources due to the optimization for undersampled data. Special processing was planned for saturated sources, but due to priorities and time constraints, was never implemented in versions prior to 990811. Design and analysis of improved saturated source position estimation is underway for V.3 of the 2MAPPS pipeline. An intermediate algorithm for saturated source position estimation was implemented and put into the pipeline processing for further testing. There was a considerable improvement in single frame positions for saturated sources. This intermediate algorithm computed the position for sources with 1 or more saturated pixels as the flux weighted centroid of the pixels over a threshold of 50 sigma. No special processing was implemented for souces near the edge of a frame, so the position was be biased in those cases.
The following plots show the difference between position estimates in the single frames between read1 and read2 for a relatively high density scan (981102n/s034). The magnitude range from approximately 4-5 to 7-8 is saturated in read2 but not read1, so the read1 positions are used as the "truth" table.
read1-read2 vs mag - centroid method
It may be seen that the positions degrade significantly over the range 8 to 4 magnitudes as the read2 data goes deeply into saturation for the pipeline processing, but not for the newly implemented centroid method. Note that these comparisons are between the single frame positions as extracted from the read1 and read2 frames, and the constant offset between the reads has not been taken out. Although this comparison is between read1 and read2 for sources saturated in read2, the algorithm has been demonstrated to preform equally well for saturated read1 sources.
000327 Update
The single frame Frexas saturated r1 extractions produced with the centroid algorithm for scans observed from 000112 to 000213 (>20000 measurements) were collected and analysed for repeatability over the 6 frames. The results are shown in this plot. The magnitude is the integrated flux over all pixels over threshold in each frame, and so is highly compressed on the bright end, the brightest source in these plots is Betelgeuse, which is actually -2.5,-3.3, & -3.5 in J, H, & Ks, respectively. This plot shows only the repeatability over 6 frames, the positions are not corrected for the r1/r2 offsets, and the magnitudes suffer from lack of zero point correction as well as errors in the integral due to saturation and thresholding. Even so, the repeatability demonstrates the potential for resonably accurate postions and photometry.
In the y residuals for Betelgeuse, the probable effect of not correcting for the frame edges may be seen in the 4 extreme measurements. In the delta magnitude plots, the residuals are seen to be worse in the 4-5 magnitude range where r1 is first going into saturation. This is due to the high snr pixels being clipped while few unsaturated pixels have significant signal-to-noise. This will be a problem regime for photometry, and may require information from r2. The current algorithm also suffers from significant variations in the lower threshold and a bug which split very bright (<-2) stars into multiple pieces.
Cong Xu has developed a magnitude estimator for bright sources using composite radial profiles fitted to the unsaturated portion of the psf for bright stars. This technique has been shown to achieve photometric accuracy of 0.2 mag or better. See Cong Xu's analyses of saturated r1 photometry compared to the CIO.
This technique is currently being implemented for the 2MAPPS V. III pipeline.
This page last updated on Nov 16, 2000.
Gene Kopan - gene at ipac.caltech.edu
