List Of Problems
Correction Algorithms
Effect On Extended Sources
This is intended to be a complete list of the various known J Band Electronic Problems seen in the data. Please let me know if I am missing any problems or key information on any of the listed problems.
We will either have to reprocess earlier data or restrict its release in the incremental release. The jumps were identified in previous scans as part of q.a., but they were just noted, with no penalty applied for them.
Davy Kirkpatrick reports:
We see the bands in all data from the south. The frequency of the banding changes over the course of the scan, and in some scans the banding level is almost imperceptible. However, if I had to guess I'd say something like 25% of all southern scans having banding similar to the one containing UGC 252.
The scan containing UGC 252 is 980416s Scan 85 Coadd 044, and has an amplitude of ± 0.25 DN.
Related webpages:
Mike Skrutskie reports:
The raw frames (flat_frames) for 980416s were very illuminating. I played them back at high speed with a skyview tak file. You can see the pickup at a low level in the upper right quadrant, although I would have expected pickup at this level to be completely benign. What happens however, is that the frequency of the pickup drifts during the scan. The bands roll south initially, then stop, then begin to roll north as their frequency increases. By the end of the scan, the frequency has increased to the point where the banding is no longer visible. The most offensive coadds must arise from the period where the drift of the bands almost exactly matches the scan rate, and thus the pickup adds coherently. All of this suggests that it will be tough to take things out in the raw frames since the signal is low and the frequency is variable.The offending quadrant is the same one that showed strange crosstalk when the Leach electronics were plugged in in the South. This may point to a hardware problem that Rae can track down in December 1998.
Gene Kopan reports:
In the J coadds for 980416s/s085, the banding amplitude appears fairly constant until the middle of ji0850127, then dies out rapidly and virtually disappears into the background level. This does seem to be quite risky to take out in the frames. I think I could make it work pretty well at high galactic latitudes when the background was flat, but it would be hard to make it robust enough to trust. A diagnostic is easier since you can tolerate a low level of false alarms. Catching subtle problems with an attempted software fix could be much more problematic. I'm not saying it's impossible, just that it could turn into a frustrating time sink.
Faint J banding is seen on the LEFT side of the array in scan 008 (subtle), 012 (subtle), 013, 026, 027, 035 (especially coadd 266), 036, 037 (wide bands that go all the way across array), 038 (has both the wide, full array banding and the left-side higher frequency stuff), 039, 040 (wide banding, but subtle), 041 (both wide [see coadd 220] and left-side banding), and 042 (subtle). Note that this is a NORTHERN night -- the previous cases of bad J banding were all in the south. Also, the list above comes from by-eye assessments of all J-band coadds in the three photometric sci blocks.Other weird quadrant effects at J in scans 021 (coadd 9), and 022 (coadd 126).
Tom Jarrett Fourier transformed the images and reports that
Yes, the banding is quite apparent. It has a period of 56" with an amplitude of nearly 0.1.
Davy reports that 980413n and 980414n show:
Although I see the occasional left-side and wide banding at J like that seen on 980412n, it is at a much reduced level and barely perceptible to the eye.
If you remove a row median from the images, you also alter or remove anything on the scale of 1/2 a frame, and you depress smaller things by fractions of the noise. To me, this seems like a bad thing to do to all the data. For the new problem [the Southern J low-frequency noise pickup], the quads seem to be affected differently, so you would have to remove the quad line median, pushing the scale of the destruction by a factor of 2. The 'bias slope residual' fix was possible with essentially no bad effects on real data, but it is a very specific fix. Also, removing a row median [for the purposes of minimizing airglow effects] will only remove that portion of OH and clouds which is correlated across the row.
Background changes more rapid than ~4' are not removed by the background fit in the galaxy processing. Residual background left in the coadds quickly become devastating to extended sources, since the flatness of the background is essential in discriminating extended sources from point sources and in obtaining accurate photometry.
When the residual background level is high, every point source appears to have extended emission around it and hence becomes an extended source. When the residual background level is low, the extended emission of every galaxy becomes lost, and every galaxy becomes a point source.
A background residual of 0.40 DN in the Atlas Images is catastrophic to extended source completeness and reliability. Before the software fix, the Northern J "reset decay pattern" background jumps caused these levels of background residuals. The Effect of J Background Jump on Galaxy Extractions shows that the completeness and reliability of extended sources has dropped to levels near 50% for each.
Analysis of Photometric Noises for 2MASS Galaxies shows that the typical error in determining the background is ~0.01-0.02 DN in the absence of additional J background noise such as these electronic problems. The noise could be as large as 0.05 DN before the total photometric noise for 2MASS galaxies is increased by 10%. Hence any additional J background noise less than 0.05 DN is not much of a problem.
Specifically, 2MASS Galaxy Photometric Error Due To J Band Noise Pickup gives the quantitative effect of a given level of additional J Band noise on the flux of the half of the 2MASS galaxies within 0.5 mag of the completeness Level 1 Specification.
