A long-standing mystery has been the high chi^2 of the band-merged Read1 position residuals vs optical catalogs. The frame-to-frame position dispersions were used to calibrate the position error models for the Read1 frame extractions. These models were verified to produce chi^2 values on the order of 1 for the frame-to-frame residuals for each band independantly. After combination of the Read1 single frame positions for each band and band-merging, the resulting positions and refined uncertainties consistently produced chi^2 values of 2 and more when compared to optical catalogs. Position errors ~100 mas were indicated instead of the predicted uncertainties of 60-70 mas.
It has been known since 1997 that Read1 frame positions have higher residuals than corresponding (R2-R1) frame positions. It has also been noted that the position residuals within each frame are highly correlated. This is thought to be due to freezing the seeing cell in the short 51ms Read1 exposures. The seeing cell size at near infrared wavelengths is on the order of the 2MASS mirror aperture. Since all three bands integrate the Read1 exposure simultaneously, this highly correlated error is also correlated between the bands as can be seen in the following plot of the per-frame position residuals for all 3 bands:
In this plot, the J residuals are plotted in blue, H in green, and Ks in red.
The seeing shape parameter for this scan was 0.950, 0.932, and 0.971 for J, H, Ks respectively.
The high degree of correlation between the residuals for all sources in all bands within each frame may be clearly seen. The in-scan residuals show larger frame-to-frame jumps than those in cross-scan, which may be due to mechanical effects.
Since the measurements for each band are assumed to be independent when band-merging the positions, but the band positions in Read1 are actually highly correlated, the position errors are underestimated by a factor of somewhere between 1 and sqrt(3).
This page last updated on Nov 1, 2002.
Gene Kopan - gene at ipac.caltech.edu
