It's long been known that the PSF estimation process for image quality assessment could be led astray by the inclusion of extractions that were not point sources. Such contamination includes:
To mitigate these problems, an update was made to the ScanFrame pipeline starting at scan 72948a. The update stipulated that during PSF construction, only those extractions matching 2MASS objects with gal_contam = 0 would be used. Such objects will be real, astrophysical sources and generally will be unresolved in WISE. Thus, they represent an appropriate subset to use in constructing the PSF. Because of the lower rate of false triggers, the improved PSF estimation allows the QA team to more efficiently identify framesets afflicted with smearing.
However, there are still instances in which this improved PSF estimation can provide misleading values:
The new metric has essentially eliminated image quality failures caused by diffraction spikes of bright stars. However, the reduction of the number of sources going into the statistics means that we have lost the ability to get adequate statistics for scan-level PSF metrics in some bins of the 5x5 subsampled array grid (for PSF monitoring toward the edges and corners of the arrays), but such metrics are rarely used. The reduction in the PSF statistics also means that the QA team now applies special scrutiny to any framesets for which the number of usable PSF stars has fallen below tolerance (currently set to 100); such framesets are usually affected by excessive moonlight only.
Last update: 24 September 2024