As with CCDs, cosmic ray (CR) hits produce unwanted signal in the
output images. Unlike standard CCD observations, however, most of the effects of CRs in NICMOS data can be eliminated during data processing thanks to the capability for multiple, non-destructive readouts in infrared arrays. As was described in Chapter 3
, the calnica
processing pipeline identifies cosmic rays in individual readouts of a MULTIACCUM image during the CRIDCALC step, and excludes data from that readout when calculating the final count rate for the affected pixel. Therefore CRIDCALC processed NICMOS images (*_cal.fits
) should be free of most or all ordinary cosmic rays. The affected pixels will be flagged in the DQ extensions of the *_ima.fits
It is worth noting that the noise level will be slightly higher for pixels
which were impacted by cosmic rays because (1) the effective exposure time for those pixels is shorter because one readout has been discarded, and (2) breaking the counts vs. time “ramp” fitting procedure into two or more pieces introduces extra statistical noise, particularly for observations limited primarily by readout noise.
Images taken in ACCUM mode have no cosmic ray processing. For
these, you must handle cosmic rays as you would for ordinary CCD images.
Occasionally NICMOS (like other instruments) will be hit by a very
strong cosmic ray, that can affect many pixels (Figure 4.15
). These will not always be effectively removed by the calnica
CRIDCALC processing. The core of the very bright CR will be flagged and removed, but there can be a large, surrounding “halo” of pixels weakly affected by CR signal that are not flagged. Also, glancing-incidence cosmic rays can leave long trails across an image, and the weaker CR pixels in the streak may not be flagged by the CRIDCALC processing. Very strong CRs can induce persistence which will alter the count rate for the rest of the MULTIACCUM sequence, resulting in a cosmic ray afterimage. If pixels reach saturation levels from a cosmic ray impact, subsequent readouts are unrecoverable.
The brightest pixels of the CR have been cleaned by CRIDCALC in calnica
, but a surrounding halo of affected pixels remains, as does a long diagonal streak of unflagged pixels from the CR’s glancing traversal of the array. Also note the vertical “Mr. Staypuft” streaks (see Section 4.8.4
) in both the left and right halves of the image, which are evidently induced by the CR impact itself.
If weakly affected pixels around bright cosmic rays are not being
flagged by CRIDCALC, you may wish to reprocess the data, reducing calnica
’s CR rejection threshold parameter crthresh
from the default (4) to a smaller value. Alternatively, you may flag the cosmic ray affected pixels manually before CRIDCALC processing. First, partially process the images through calnica
excluding the CRIDCALC stage (i.e. setting cridcalc to omit). Next, in the resulting *_ima.fits
image, edit the DQ extension of the first IMSET affected by the cosmic ray, setting the affected pixels to the cosmic ray flag value 512 (see Table 2.3
). Finally, set cridcalc to perform in the *_ima.fits
file and complete the calnica
processing using the modified *_ima.fits