Each telescope is a three-mirror anastigmat featuring a rectangular aperture2.1of , an inter-mirror distance of about 3 m and a focal length of 50 m. The image scale on the focal plane is thus of about , while the instrument's Airy Disk is an ellipse of axes at and at . The aperture is elongated in the scan direction, so as to provide the narrowest PSF in the measurement direction while being compatible with the volume of the spacecraft and the optical quality of the mirrors. The large primary and tertiary mirrors are polished to rms, while the smaller secondary mirrors are polished to rms, yielding a diffraction limited performance over the whole field of view.
The focal plane of each telescope is basically a rectangular mosaic made of more than 300 of CCDs, giving a field of view of . A rectangular pixel size of or was chosen in order to match the Airy Disk shape and thus providing an higher resolution in the scan direction. The large focal length allows a proper sampling of the diffraction pattern with about 4 pixels along scan and 3 pixels across scan covering the Airy Disk. CCDs of two different sizes are used in order to increase redundancy in some key areas of the focal plane such as the sky mapper and the overlapping regions. Smaller CCDs have a size of , while bigger CCDs are the same size along scan and twice as large across scan, giving a size of . An observed object follows a nearly horizontal line on the focal plane with a speed given by the spinning period, and therefore successively crosses all the columns of CCDs. With the 3-hour spinning period provided by the scanning law the object has an along-scan speed of , corresponding to about 3200 pixels/s or 0.31 ms/pixel. Due to the high speed, the charges accumulated in the CCD pixels cannot be read out as it is done with conventional imaging telescopes, but a dedicated integration technique must be used. In the case of GAIA, the CCDs will be operated in Time Delay Integration (TDI), a concept introduced for an astrometric satellite by [Høg 1993]. The idea is to let the integration process follow the image while it is moving across the CCD. In practice, every 0.31 ms, i.e. every time the image has moved of one pixel along scan, all charges are quickly shifted by one pixel in the scan direction. The readout of the accumulated charges takes place at the serial register at the ``end'' of each CCD. The image is thus integrated over the entire crossing of each CCD, leading to an exposure time of about 0.86 s per CCD per scan. The main drawback of this technique is the additional smearing of the image due to the charge shift and to the slow across-scan motion of the objects, which together cause an appreciable but acceptable loss of resolution. On the other hand, the loss of resolution due to the non optimal charge transfer efficiency is expected to be negligible.
The focal plane layout is represented in Figure 2.7. The 25 columns of detectors covering the focal plane are functionally grouped in four parts: the Astrometric Sky Mapper (ASM, 4 columns), the Astrometric Field (AF, 16 columns), the Photometric Sky Mapper2.2(PSM, 1 column) and the Broad-Band Photometer (BBP, 4 columns). A thorough description of the functions fulfilled by the different parts of the focal plane is given in Section 5.1. In brief, the ASM is used for object detection, the AF for multiple multi-epoch astrometric measurements, the PSM to generate a 1 arcsec radius high-resolution map around each detected star and the BBP for multi-color and multi-epoch broad-band photometry.
The ASM, the AF and the PSM work without filters in a very broad band, having a response curve defined by the telescope transmittance and the Quantum Efficiency (QE) of the presently agreed-upon CCD, the so called CCD#1B. The resulting response curve extends in the range of wavelengths 250-1050 nm, and the zero-point of the magnitude scale, the so called (GAIA) magnitude, is such that for most stellar types the magnitude has a value which is intermediate between and . The BBP columns work in four different broad bands defined by the product of the aforementioned global response curve and the response curve of four filters.
The photometric system ([Høg, Knude and Straizys 1999]) adopted in this study consists of five passbands of about 100 nm width. A rectangular response and a peak transmission of 0.90 is assumed for all filters. Roughly speaking, the bands closely resemble the bands, while the bands together cover a waveband with the same center as the band. The QE curve of the is shown in Figure 2.8, in which the normalized response curves for the Asiago Photometric System ([Munari 1999]) have been preferred for illustrative purposes to the corresponding curves for the system. The specifications of the two photometric systems are given in Table 2.1.