5.1 Detection and Observation

During the scientific operations, the GAIA satellite will continuously spin about its symmetry axis, and the charges contained in the CCD pixels will correspondingly be shifted along-scan to integrate the image for a sufficiently long exposure time. This observing strategy, while allows the coverage of the whole sky for on average 150 times during the mission, requires the implementation of a dedicated CCD readout process. The CCDs covering the focal planes of the Astros, in fact, contain plenty of data that must be readout approximately every 0.86 s, in order to catch up with the scanning of the satellite and the shifting of the charges along the CCDs. Should every bit of this huge amount of information be readout, this would yield an unbelievably high reading frequency and telemetry rate. Besides, the levels of readnoise^5.1following from the high reading frequency would seriously compromise the quality of most data. This called for a strategy to discriminate the sky regions containing scientifically interesting objects from those containing only noise or other uninteresting features in an automated manner, and to restrict the readout to the former.

Two possibilities for defining the interesting sky areas have been considered. The use of an input catalogue of about 100 million stars selected for their astrometric and astrophysical interest, would allow to limit the readout to the areas around these stars. But no survey exists that could be used as basis for a meaningful selection, and the construction of such a catalogue from scratch would be a very time-consuming and expensive undertaking.

The other possibility, the one finally chosen, is to detect stars as they enter the field of view, determine their position, magnitude and signal-to-noise ratio, and, if the latter exceeds a certain limit, to collect data around such stars during the remaining part of the field crossing. It has been shown that the amount of data can thus be reduced by at least a factor thousand. A further advantage of this approach is a clearer characterization of the observational selection effects, which is extremely important in evaluating the completeness and properties of the observed sample.

In this context, it is important to emphasize the distinction between the detection and the observation of an object with GAIA. Whether an object will be observed or not is determined only by its detection as it enters the field of view, not by any prior knowledge of its position, even if it has been previously observed by the satellite. Stars near the detection limit will not be generally detected and therefore observed during all scans. The on-ground database will however be able to instantly access all available observations of any required star or sky area for the purposes of data reduction.

The presently agreed-upon arrangement of the Astros focal plane, which was shown in Figure 2.7, resulted from detailed and extensive simulation of observations, and closely reflects the observing strategy outlined above. The CCD columns can be logically divided into four parts, whose roles can be described as follows:

• Astrometric Sky Mapper (ASM0-3): the main function of the ASM is object detection by real-time on-board data analysis. ASM0 and ASM1 are used for the detection and position measurement of bright and faint stars, respectively. ASM2 is redundant, i.e. is not used in normal operations but is kept as a replacement if another should fail. ASM3 is used to confirm the object detection, to refine the measurement of its position and to determine the satellite scan velocity. The position and velocity measurements, in particular, are used to propagate the measurement window through the remaining part of the field of view.
• Astrometric Field (AF1-16): the AF is used for astrometric and photometric observations. Its 16 CCD columns are used to measure astrometrically and photometrically (in the $G$ band) the detected stars 16 times.
• Photometric Sky Mapper (PSM): The single CCD column of the PSM is used to map the surroundings of the detected stars within a region of radius 1 arcsec, in order to correct the photometric observations carried out in the AF for the effect of disturbing stars.
• Broad-Band Photometer (BBP1-4(5)): each of the 4-5 CCD columns of the BBP is used to measure photometrically the detected stars in a different photometric band. Note that the other parts of the field of view all work in the $G$ band defined by the telescope transmittance and the CCD quantum efficiency.

The general philosophy underlying the CCD readout and the data transmission must be that of maximizing the scientific content per readout/transmitted bit, taking into account that a large CCD binning size implies a reduction in the readnoise, which is the dominant noise source for faint objects, far larger than the sky background. Therefore, over the four different parts of the field of view, and, in the case of the ASM, even over the different columns of the same part, the CCDs are readout with a different binning matched to the specific scientific purpose, which also determines the area of the sky region around each detected object from which data are transmitted to the ground. As far as nomenclature is concerned, the CCD elementary binnning region is called a sample^5.2, while the sky region that is observed around each detected object is called a patch.

The optimization of the all-mission astrometric and photometric performance with respect to sampling has lead to the rectangular, slightly elongated shape of ASM samples, and to much more elongated, essentially one-dimensional, samples in the other parts of the focal plane. Clearly, this different choice is due to the need of performing two-dimensional position and velocity measurements at each scan in the ASM, while one-dimensional single-scan measurements in the other CCDs suffice to reconstruct the positions and motions of the observed objects.

The presently planned sampling scheme of GAIA instruments is illustrated in detail in Figure 5.1. Note, in particular, that the sample size chosen for the BBP of the two Astros is different, being $1\times8$ pixels in Astro 1 and $6\times8$ pixels in Astro 2. With this choice the Astro 1 yields a better photometric accuracy for bright stars, wheras the Astro 2 is more accurate for faint stars. Simulations have shown that with such an observing strategy stars of,e.g., $I=20$ can be detected with a probability of 0.9.

Figure 5.1: The sampling scheme in GAIA instruments. Sample and patch sizes are indicated, together with the expected readnoise levels. The Airy Disk of the Astros is shown as a solid ellipse, while the dashed ellipse shows the elongation due to the maximum value of the across-scan motion of the image during the 0.86 s integration time. From [Høg 1999]. Courtesy of Erik Høg, Copenhagen University Observatory.

Another stringent requirement on the CCD readout process is that, in order to ensure the greatest thermal stability, the reading frequency of all the CCDs must be kept constant. Roughly speaking, the reading frequency (or telemetry rate) can be written as the product of two factors, namely the readout (or transmitted) data per detected object and the number of detected objects per unit time. The first factor depends on the sample and patch sizes, actually on the ratio between patch and sample size, which vary greatly depending on the column of CCDs under consideration. The second factor depends on the number of objects brighter than the detection limit per unit solid angle, which in turn is a strong function of the Galactic coordinates of the instantaneous field of view, being very high near the Galactic plane and relatively low at high Galactic latitudes, with smaller variations on very short space scales. Consequently, in the course of the mission the reading frequency of useful data vary greatly on short time scales, and therefore most of the time some dummy samples not belonging to any object are readout, but not used further, to ensure that the total reading frequency is kept constant. The presently planned value of the total reading frequency is set by the requirement of observing up to 2800000 stars/deg$^2$. Only a small fraction of the sky has a such a high star density, and it has been suggested that this requirement could be relaxed in order to obtain a lower readnoise per unit area, which would result in a higher accuracy. The readnoise could be very effectively decreased by the use of a total reading frequency matched to the number of samples to be read, but at the cost of complexity in the thermal control.

Once the samples from the Astros have been transmitted to the ground, they can be analyzed in various ways to obtain astrometry and/or photometry^5.3. The samples belonging to a patch can, e.g., be analyzed alone to give epoch astrometry and photometry. All the patches belonging to the same sky region may then be analyzed together to get all-mission average measurements of the astrometric parameters. When all-mission averages of the astrometric parameters are known, it is possible to return to the single patches in order to derive more accurate epoch and all-mission photometry. Generally speaking, the derivation of the astrometric parameters will precede the photometric analysis. Thus, the astrometric parameters and the energy flux of the objects in several colors may be derived from all the patches covering the same sky region.