CCDs in space

Space-based CCDs have a number of special problems associated with them that are often not considered for ground-based systems. Once launched, human intervention is unlikely and the CCD and instrument package can never be retrieved for fault correction or calibration purposes. Even simple procedures, such as bias calibration, take on new meaning as CCD evolution or changes in the gain or other CCD electronics mean new calibration images are needed. Damage to the array (see Section 7.2), or the possibility that the primary circuits fail and the backup electronics or even a different clocking scheme must be used, means that new calibration images must be produced. Also, each observer does not have the ability to obtain all the needed calibration data and the project must provide the finest and most up-to-date calibration images for each CCD, instrument, and mode of operation. All issues have to be thought out completely prior to launch or dealt with through analysis of downloaded data during the mission.

One such example of a significant change in CCD operation is provided by the Hubble WFPC2 instrument (Holtzman et al., 1995b). After operating in space for only about three months, it was noticed that the CCDs developed an odd sort of CTE effect. The effect caused stars to appear fainter if imaged in higher numbered rows. The apparent cause was the development of a large number of traps within the CCDs not seen during preflight tests. Photometric gradients of 10-15% were present along CCD columns and, even worse, the effect was highly dependent on the brightness of the imaged star, being only about 5% for bright stars.

Using ground-based laboratory tests with similar electronics and CCDs, it was determined that changing the operating temperature from -76° C to -88° C would cause a sharp decrease in the CTE effect. Such a change caused the CTE variations to almost disappear, leaving only a 3-4% gradient. A further temperature decrease would probably improve the situation but inflight hardware did not allow the CCDs to be operated at colder levels. Thus considerable effort has been put into the development of a semi-empirical software model that can be applied to data obtained with the WFPC2 in order to correct for the remaining effect (Holtzman et al., 1995a; Whitmore & Heyer, 1998). A number of the CCDs in HST instruments and those in the

Chandra X-ray observatory have shown long-term degradation in their CTE performance due to the radiation environment of the telescopes' orbit. For example, the STIS CCDs have changed from a CTE of 0.999 999 to 0.999 91 since launch (Kimble et al., 2000).

One consequence of the CCD operating temperature being lowered in the WFPC2 was decreased dark current. However, on-orbit hot pixel development was greater than expected with many of these hot pixels "fixing" themselves after dewar warming (Section 7.2). Calibration dark frames are therefore required often to monitor the dark current and to provide the best dark frames to use given any set of observational circumstances. Hot pixels are especially important to understand in space-based CCD imagery as the very small PSF of imaged scenes and the appearance of numerous cosmic rays with a plethora of shapes, including single pixel events, must be distinguished from the collected flux of interest.

We alluded above to the importance of cosmic ray identification in order to avoid misinterpretation of imaged scenes. From a sample of 2000-second dark images taken with the WFPC2 it was found that 5-10% of the cosmic ray events were single pixel events of 5 sigma or greater above the bias level. Fully one half or more of these events showed consistent pixel positions from frame to frame and thus could not be identified with true cosmic rays or local radioactivity from the dewar and surroundings. Typical signal levels for true single pixel cosmic ray events were near 200 electrons while multiple events peaked near 700 electrons (Holtzman et al., 1995b). Multiple pixel cosmic ray hits (averaging seven affected pixels per event) are much more common than single pixel events, and a rate of almost two events per CCD per second was observed.

CCD dewars, once sealed, evacuated, and chilled, are often seen to produce contaminants owing to outgassing of grease or other coatings used in their construction. When at operating temperatures of -80° C or so, the dewar window is a good site for condensation of such contaminants. These small particles of material are very good absorbers of light, particularly UV and visible radiation, because of their characteristic sizes. A likely cause of the contamination is C, O, and F atoms that often form a thin layer on the dewar window or instrument filters quickly and then increase this layer slowly with time. Bake-out procedures have been modeled as a possible method to reduce the thickness of these layers (Plucinsky et al., 2004) specific to the ACIS CCDs on the Chandra X-ray observatory.

One simple calibration test that allows monitoring of this effect is to obtain fairly regular observations of a bright UV star. If the dewar window does indeed get fogged with material, careful measurements of the UV throughput of the observed flux will show a slow degradation. Even in the best space-based instruments, small amounts of material outgas, and after several weeks UV performance can be noticeably lower. One solution that seems to work, at least for the Hubble Space Telescope WFPC2 CCDs and the Chandra X-ray observatory ACIS CCD imager (and for general observatory dewars), is to warm the dewar up to allow for thermal desorption. The WFPC2 CCDs were warmed to near 20° C for about 6 hours approximately every month. In a typical observatory dewar after warm up, one can attach a vacuum pump and pump out the now non-frozen water and other contaminants, then recool the device.

Flat fields, as we have discussed before, are very important to have in one's calibration toolkit. Once in orbit, either as a satellite or as a spacecraft destined for another world, the CCDs aboard generally have little ability to obtain flat field calibration images. High S/N flats made prior to launch in the laboratory are often the best available. These usually provide overall correction to 5% or a bit better, but small effects, such as illumination or instrument changes, limit the accuracy of the correction. Sometimes, the space-based CCD has significant changes, and large corrections are needed or new flats have to be generated in some manner.

The original WFPC camera aboard Hubble could obtain on-orbit flats through observation of the bright earth (Holtzman, 1990; Faber & Westphal, 1991). These were not elegant flats, having streaks and nonuniformities, but were all that was available. WFPC2 used Loral CCDs, which have an increased stability over the original TI CCDs, allowing preflight laboratory flats to work very well, even after the reduction in operating temperature as discussed above. Numerous other small effects, such as color dependence, radiation damage, hot pixels, CCD illumination, and optical distortions seen in the on-orbit WFPC2 flats are discussed in detail in Holtzman et al. (1995b). The effects of flat fielding, CTE, and the other issues discussed above on the photometric performance of the Hubble WFPC2 are described in Faber & Westphal (1991), Holtzman et al. (1995a), and Whitmore & Heyer (1998).

The Galileo spacecraft certainly provided impressive imagery of the planet Jupiter and its satellites and was one of the first public CCD cameras to be launched into space. Its CCD camera is described in detail in Belton et al. (1992) and can be used as an example of the details of space-based observations, their calibrations, properties, and difficulties. CCD and instrument stability and processes for their calibration after launch are major effects to consider as well as proper treatment of the photometric calibration images in lieu of the much reduced PSF.

The solid-state imager (SSI) aboard Galileo consisted of a single 800 x 800 TI CCD with a read noise of 40 electrons, gains of 38 to 380 electrons per DN, and a pixel size of 15 microns yielding 2.1 arcsec per pixel. The SSI, like the WFPC2, developed a CTE problem after about 8-12 months in space. Detailed study of SSI images taken during periods of cruise science (Howell & Merline, 1991) revealed that the CTE problem resulted in a readout tail containing 400 electrons, independent of the brightness of an imaged star or its location within the CCD. The cause was attributed to a trap, not in the active CCD array, but in the output register. Radiation damage (see next section) was the most likely cause. Due to the constant number of trapped electrons, photometric correction was possible to a high degree of accuracy.

Point sources imaged in space are free from the blurring effects of the Earth's atmosphere and have a very small PSF compared with those commonly obtained with ground-based telescopes. A theoretical diffraction-limited image formed through a circular open aperture will have a FWHM (of the Airy disk) in radians of

1.03A

where A is the wavelength of observation and D is the diameter of the aperture (Born & Wolf, 1959). Note that if we were to use the radius of the first Airy disk dark ring as our definition of image size, we would have the traditional formula

1.22A

Figure 7.2 shows theoretical Airy disk PSFs expected to be imaged by the SSI at three representative wavelengths and five different possible slight de-focus values.

The FWHM of the SSI images (being obtained without any atmospheric or other seeing effects) were predicted to be about 0.55 arcsec at 4000 A and 1.2 arcsec at 9000 A. These PSF sizes correspond to 0.25 and 0.6 pixels respectively, making the SSI images severely undersampled (r ~ 0.2). This level of undersampling makes it impossible to directly determine the true FWHM or profile shape of a PSF. Using multiple images with slight offsets, images containing multiple stars with different pixel grid placements, and model CCD images, one can reconstruct the true PSF imaged by an under-sampled space-based CCD camera. In the SSI case, the PSF was found to be slightly larger than predicted and attributed to a slight camera focus problem.

As we have seen, undersampled images will lead to astrometric and photometric error, as the lack of a well-sampled PSF makes it hard to determine the true image center or the actual flux contained in the image. For the SSI,

Fig. 7.2. Modeled Airy disk patterns imaged by the Galileo SSI. The top panel shows the calculated PSFs as would be seen under very well-sampled conditions while the bottom panel shows the same PSFs as they would appear when imaged by the SSI. The severe pixelization of the PSFs is apparent. The rows are for 7000, 5500, and 4000 A (top to bottom) and the five columns are (left to right) de-focus values for the SSI camera in mm. From Howell & Merline (1991).

Fig. 7.2. Modeled Airy disk patterns imaged by the Galileo SSI. The top panel shows the calculated PSFs as would be seen under very well-sampled conditions while the bottom panel shows the same PSFs as they would appear when imaged by the SSI. The severe pixelization of the PSFs is apparent. The rows are for 7000, 5500, and 4000 A (top to bottom) and the five columns are (left to right) de-focus values for the SSI camera in mm. From Howell & Merline (1991).

astrometric error amounted to about 0.8 arcsec even for bright stars, or about half a pixel. Observations of bright guide stars are a common occurrence for spacecraft and are used for navigation and course correction. Large astrometric uncertainties are hazardous and can lead to spacecraft orbital trajectories with inaccurate pointings, having the potential of producing spacecraft course corrections that could cause it to miss a target or, even worse, come too close. In the Galileo case, it was determined that a large number of guide star images was needed and careful analysis of these could be used to determine the path and navigation of the spacecraft within acceptable limits.

Photometrically, the nature of the undersampling manifests itself in two ways. First is the way in which one extracts the data and how a flux value is assigned to it; second is the effect of digitization noise, which is large for the SSI. Figure 7.3 illustrates the first of these issues by presenting SSI data for a bright star. Because of the nature of the PSFs imaged with the SSI,

Radius (pixels)

Fig. 7.3. Radial profile plot of a bright star imaged by the Galileo SSI. The plus signs are the actual CCD DN values (for G = 380e-/DN) and HWHM and r correspond to predicted values for an Airy disk imaged at 5500 A. Note that an approximation as a Gaussian profile is a poor representation of the actual PSF but the determined EHWHM for a single measurement is not far off. From Howell & Merline (1991).

Radius (pixels)

Fig. 7.3. Radial profile plot of a bright star imaged by the Galileo SSI. The plus signs are the actual CCD DN values (for G = 380e-/DN) and HWHM and r correspond to predicted values for an Airy disk imaged at 5500 A. Note that an approximation as a Gaussian profile is a poor representation of the actual PSF but the determined EHWHM for a single measurement is not far off. From Howell & Merline (1991).

one pixel (plus sign at r = 0.25) contains much more flux than any of the remaining ones. A standard Gaussian fit to these data (in this case made by IRAF) is seen to provide a complete misrepresentation of the image profile. Imagine the photometric error one would introduce by assumption of this type of profile and use of its shape as an indication of the total counts observed for this star. The effective FWHM (EFWHM) is defined as the apparent PSF width as determinable from a single undersampled image of a star. We see here that the EHWHM is 0.7 pixels, compared with the expected value (at 5500 A) of 0.55. The digitization effect present at the highest SSI gain setting leads to an uncertainty of ±379 electrons per DN. The above effects combined lead to an overall relative photometric uncertainty of 5-10% and an absolute spectrophotometric uncertainty of 10-30% for SSI data. These are higher than the 2-5% uncertainties quoted for the WFPC2 camera and are directly in proportion to the greater undersampling and higher CCD gain values used in the Galileo SSI.

Further readings concerning the special conditions and circumstances of CCDs when used for space-based observations can be found with a quick search of the websites of the Hubble Space Telescope and other satellite and spacecraft observatories. Access to numerous internal technical, engineering, and calibration reports is given as well as literature articles containing applications of the findings to astrophysical objects.

Was this article helpful?

0 0
Telescopes Mastery

Telescopes Mastery

Through this ebook, you are going to learn what you will need to know all about the telescopes that can provide a fun and rewarding hobby for you and your family!

Get My Free Ebook


Post a comment