reduced, with a consequent decrease of the read-out 
noise and of the CCD read-out time. Of course, the 
reduced read-out noise is advantageous for the 
detection of fainter stars. 
In the Astrometry area, the nominal TDI shifting period 
(with angular velocity of 120 as/sec and pixel size of 4 
mm) is Tshitt * 172 ps. 
The Photometry area and the Star Mapper are covered 
by mosaics of 40 and 28 CCDs respectively. In these 
areas, the sampling requirements are more relaxed. A 
typical sampling of a star image of 3 x 3 pixels is 
sufficient to determine photocenter coordinates. The 
pixel size is 60pm x 120pm, corresponding to a TDI 
shifting period of T S hift * 2580 ps. 
The major characteristics of the CCD for all the areas 
are summarized in table 6 and 7, the integration time 
reported is referred to an elementary exposure, that is 
the exposure on a single CCD during the star transit on 
the detector. 
Table 6: CCD characteristics 
Astrometry 
Area 
Photometry 
Area 
Star Mapper 
CCD linear 
size 
23.3x70 mm 
35 x 70mm 
35 x 70mm 
Pixel size 
4 pm x 50 
pm (linear) 
60pmx120 
pm (linear) 
60pmx120 
pm (linear) 
Number of 
pixel in CCD 
-5800x1400 
580 x 580 
580 x 580 
Integration 
time 
1 sec 
1.5 sec 
1.5 sec 
Table 7: CCD sensor requirements for the astrometry area 
Image 
sampling 
> 3 pixel/fringe period 
Quantum 
efficiency 
> 60% 
in the X = 750 ±100 nm observation 
band 
Read-out- 
noise 
< 3 e 
Charge 
transfer 
inefficiency 
<10' 6 
3. METROLOGY 
The metrology for the GAIA interferometric instrument 
is one of the most crucial issues to the success of the 
mission. 
The optical part of the instrument must be kept stable 
essentially for two reasons: formation and persistence 
of the interference fringes, and stabilization of the basic 
angle (Cesare, 1998). 
Concerning the former, the fringe visibility must be 
always close to its nominal value (visibility loss < 5%) 
across the astrometric field. The second reason is 
nevertheless crucial, in fact while the basic angle 
between two lines of sight needs to be controlled within 
5 pas, the stability of the relative position of star images 
in two different FOVs is relaxed to 10 pas to take in to 
account geometric field distorsion variability, in the time 
scale from 0.75 sec (minimum integration time of a star 
image on the astrometric field) to 3 hours (great circle 
scan period). 
The most critical elements to be stabilised are the 
primary mirrors of the optical interferometer which can 
be subjected to rigid body motions like translations 
(tolerances: 5 nm) and rotations (tolerances: 1.8 nrad) 
and the beam combiner mirrors (tolerances: 17 prad in 
rotation, corresponding to a linear translation of 6 pm). 
The mirror shape variations are less critical, in fact they 
have smaller effects on the fringe visibility loss, and 
hence on the image position shift, and can be passively 
controlled (i.e. by passive thermal stabilisation). 
The rigid-body relative movements of the most critical 
mirrors (beam combiner and primary mirrors of the 
optical interferometer) are monitored and controlled by 
means of an Optics Active Control System composed 
by 15 laser interferometers (each measuring the 
distance between two reference markers) and 5 tip-tilt 
mechanisms, each controlling the mirror in three 
different directions: one translation along the mirror 
longitudinal axis, and two rotations - cx-tilt, P-tilt - of the 
plane normal to this axis. 
Three markers are placed on each of the primary 
mirrors (M1*), 6 on the mirror M2, and 18 are distributed 
on the beam combiner mirrors (figure 8). The markers 
are attached to the mirrors through optical contact. 
The laser interferometers monitor the distance variation 
between reference markers (caused by rigid-body 
movements of the mirrors) and the tip-tilt mechanism 
moves the “active mirrors” to compensate the distance 
variations. Moreover, several tests and simulations 
have demonstrated that, as a consequence of the 
distance control between the selected reference points, 
also the fringe contrast, the star relative position and 
the basic angle are also kept under control. 
Error analyses have shown that the measures of both 
absolute and relative distances must be performed by 
laser metrology with the following accuracy:5s/s = 4-1(r 
1a (relative error for the absolute distance 
measurements), and 5s/s = 8-10‘ 12 1a (relative error on 
distance variation measurement). In order to attain 
these results the laser frequency variation must be 
stable at 8v/v < 8-10' 12 1a over 0.75 s + 3 h time scales. 
Figure 8: Position of the reference markers (L, Mi, Rj, Si,) 
on the optical interferometer mirrors 
150