|Space Telescope Science Institute|
|WFC3 Instrument Handbook|
Like the UVIS channel, the IR channel uses 16-bit Analog to Digital Converters (ADCs), providing a digital output signal in the range between 0 and 65,535 data numbers (DNs). The default gain setting for the IR channel is 2.5 electrons/DN and is the only one offered to observers. Note that this setting results in a measured gain of ~2.4 electrons/DN; see Table 5.1.The default gain is compatible with the full-well saturation level of the IR channel, which is about 78,000 electrons (~33,000 DN at the default gain), and with the detector readout noise, of order 20-22 electrons per correlated double sampling.Cosmic rays affect the image quality. On-orbit measurements indicate that cosmic ray events occur at a rate of 11 ± 1.5 CR/s for WFC3/IR. The use of MULTIACCUM mode makes it possible to filter out cosmic rays because it provides a series of intermediate non-destructive reads. The intermediate reads are used by the WFC3 data pipeline to identify cosmic ray hits, similar to the use of CR-SPLITs in CCD observations. AstroDrizzle also checks for and attempts to remove cosmic rays from the “drizzled” images when there are multiple (usually dithered) exposures of the same field. (See the DrizzlePac documentation.)Passages through the South Atlantic Anomaly (SAA) cause the highest number of cosmic ray hits. When the HST is within the predefined SAA exclusion zone, IR observations are not normally taken and the detector is set to auto-flush mode to minimize the effects of SAA passage on instrument performance. Unlike NICMOS, where the detector electronics had to be switched off during SAA passage, it is possible to perform time-critical observations in the SAA with WFC3/IR.Snowballs (named for their fuzzy appearance) are transient extended sources that appear in IR exposures at the rate of about one per half hour. They may be caused by radioactivity in the detector or bonding material (WFC3 ISR 2009-44). A snowball affects between 11 and 34 pixels and contains between 200,000 and 900,000 e-. (See WFC3 ISR 2009-43.) The energy is deposited in the pixels instantaneously, so snowballs can be removed via up-the-ramp fitting, like cosmic rays.7.9.3 On-Orbit DegradationUnlike the CCDs, minimal cosmic-ray damage to the IR detectors is anticipated. During ground testing using a particle accelerator, the WFC3/IR arrays were subjected to radiation doses much higher than expected in their entire orbital lifetime, without sustaining significant long-term damage or measurable degradation in QE. Searches for the development of new bad pixels are conducted as part of the regular calibration program; the number is growing slowly if at all (WFC3 ISR 2010-13, WFC3 ISR 2012-10).7.9.4 Image PersistenceThe IR detector exhibits image persistence, particularly during and following observations of targets that saturate the detector by more than the pixel full well depth. Characterization of image persistence is ongoing. The amount of image persistence depends on the brightness of the source and possibly the amount of time the light from a bright source is incident on the detector. Image persistence has been observed both within a set of dithered exposures in a single orbit and in observations where the target observed in a previous orbit was particularly bright. Persistence appears to decay roughly as a power law with time, at least for delay times greater than about 100 s. All bright sources (close to full well or greater) exhibit some persistence. The amount of persistence for a given pixel increases very slowly if at all after saturation is reached, and therefore the primary affect of observing a star that is many times saturated is to increase the number of pixels in the psf of a star that reach saturation. Consequently, the persistence image of stars are usually much less sharp than normal images of stars. Many of the characteristics of the persistence are explained by the model of persistence described by Smith (SPIE 7021-22, Marseille 2008-06-24) for HgCd detectors.The most common form of image persistence is due to bright stars in the field, which saturate or nearly saturate an observation. An example of this is shown in Figure 7.8, which is one of many dark exposures that have been obtained with WFC3/IR. In the previous 0.7 to 3 hours, a starfield containing a number of bright stars had been observed in a linear dither pattern. Residual images of the bright stars appear in the dark frame. The count rates observed an orbit later from the most saturated stars are as large as 200 e–/s in the flat field image. Close inspection of the images shows that the shape of the residual images is less peaked than those of real stars, reflecting the fact that the persistence image is not a simple fraction of the initial count rate.Figure 7.8: A dark exposure with the WFC3 IR channel affected by persistence images. The previous set of exposures with WFC3 IR contained several stars that were highly saturated. The linear dither pattern is clearly visible.Figure 7.9 shows a less common problem in which a bright extended source (a globular cluster) was observed in an orbit prior to the observation shown. Here the persistence image extends over much of the field, and may compromise the science from the program in question. The highly saturated observation was from a prior observer’s program.Figure 7.9: A observation of a fairly sparse field observed with WFC3 IR which shows the residual image of a globular cluster. The series of dithered exposures of the globular cluster had been obtained about 2 hours prior to this observation. The displayed image has been stretched to emphasize the residual image which has peak typical count rates of 0.4 electrons/sec, considerably less than 1% of the fluxes in the original exposures. The exposures of the globular cluster had numerous stars with typical fluxes corresponding to 1000 e–/s.Persistence as prominent as shown in the examples above is infrequent, affecting about 5 observers in Cycle 17. STScI does try to identify observations that are likely to cause persistence like these examples, and as part of the planning process, STScI does attempt to prevent a close follow-on observation from taking place. This is one reason that the number of really bad cases of persistence was as small as indicated. We will continue to try to improve this in Cycle 18 and beyond, but it is not feasible to spread out IR observations so far that there will be no persistence in any observers images. When an image is corrupted by persistence due to an earlier observation to the extent that it fundamentally compromises the science proposed, observers can request to have the observation repeated. For most observers, persistence from previous observation, if it does exist at all, is a nuisance but does not compromise the science one would expect to obtain from the IR channel on WFC3. For these datasets, it is likely that the persistence signal can be reduced by a factor of 10 by post-processing, as described by Long, et al. at the 2010 STScI Calibration Workshop.If observers suspect that prior saturated observations may have left residual images in their data, the HST History Search can be used to determine the timeline of observations just prior to the compromised data set. If there are no IR observations within 4 hours of your exposures, it is very unlikely that your IR images contain significant amounts of persistence. Additional information about how to identify persistence is contained in the HST Data Handbook.Observers need to consider persistence in the planning of their observations, especially if dither offsets are much larger than the psf or if mosaics are being constructed within a visit. Planning observations of dense fields, globular clusters, or very bright star formation regions may be complicated in instances where precise photometery or very large dynamic range are required. Trade-offs must be evaluated between dithering and persistence, depending on the effectiveness of the persistence correction tools currently being developed. Specifically, observers should take care that bright or saturated sources within a field are not dithered across the same parts of the detector that record the highest science priority areas of the field. It is important to remember that fairly short exposures (100 s) of relatively bright (18th mag) stars observed through the broad-band filters saturate pixels in the IR array and so there will often be “self-induced” persistence within visits. Various tools are described in Appendix D:Bright-Object Constraints and Image Persistence to estimate which regions of an image are likely to cause persistence.
7.9.5 The IR BackgroundIn space, the dominant sources of background radiation are zodiacal light and earthshine at shorter IR wavelengths, and the telescope thermal emission at longer wavelengths. For HST, the sum of these two background components has a minimum at about 1600 nm (roughly lying in the H band).The surface brightness of zodiacal light is a function of the ecliptic latitude and longitude, and reaches a minimum ~120 degrees from the sun, i.e., near, but not at, the ecliptic poles increasing again to 180 degrees (anti-solar direction).Figure 7.10 shows the observed background levels in the first few months of WFC3’s operation (green), compared with values predicted from known instrument sensitivities and expected levels of zodiacal light. Additional analysis of the IR background is underway, as described by Pirzkal at the 2010 STScI Calibration Workshop.Figure 7.10: Infrared background levels for WFC3. The average, low and high zodiac points show the synphot predictions for three levels of zodicial light and no earth-shine. The observations selected include calibration and GO images of empty or sparsely populated fields, giving an accurate representation of early WFC3 observations but not covering all background conditions. Observational means (green points), ranges (error bars), and exposure counts are plotted.For pointings very close to the bright Earth limb, the zodiacal background may be exceeded by earth-shine. The brightness of the earth-shine falls very rapidly with increasing angle from the Earth’s limb (due to the effectiveness of the HST baffles), and for most observations only a few minutes at the beginning and end of the target visibility period are significantly affected. The major exception to this behavior is a target in the continuous viewing zone (CVZ). Such targets will always be rather close to the Earth's limb, and so can sometimes see an elevated background for part of the orbit, even at shorter wavelengths where zodiacal emission ordinarily dominates.For targets faint enough that the background level is expected to be much brighter than the target, the observer has two options when crafting the Phase II proposal:
(1) specify a non-standard Bright Earth Avoidance (BEA) angle, which increases the angle from the Earth's limb from 20° to 25° (note that this is an available mode and must be specially requested through a Contact Scientist), or (2) specify the LOW-SKY option, which restricts observations to targets more than 40° away from the Earth's limb and restricts scheduling to times when the zodiacal background is no greater than 30% above the minimum achievable level. The second option decreases the available observing (visibility) time during each orbit and implies scheduling constraints.7.9.6 BlobsImage blemishes peculiar to the WFC3-IR detector have been described in Section 5.7.7. Other spots of reduced sensitivity in IR images, dubbed “blobs”, are caused by reduced reflectivity of the Channel Select Mechanism mirror. Blobs were first noticed in IR images shortly after WFC3 was installed on HST. Their number increased to ~25 in early 2010, stabilized, then began to increase again in the latter part of 2011. They have a measured half-light radius of 10-15 pixels and absorb up to 15% of the incoming light at their centers. They are described in detail in WFC3 ISR 2010-06. Those appearing through cycle 19 have been flagged in the IR bad pixel table (WFC3 ISR 2012-10). Appropriate dither strategies to mitigate the effects of blobs and other artifacts in IR images are described in WFC3 ISR 2010-09.7.9.7 Optical AnomaliesAnomalous features are found in some IR detector images. For example, the optical system may cause stray light from sources outside the detector FOV to be scattered into images. Scattered earthlight can greatly increase the background on part of the detector when HST is pointing near the bright earth limb. (See Section 6.10 in the WFC3 Data Handbook.)