Astrometric meaning and interpretation of high-precision time delay integration CCD data

We investigate the astrometric content of CCD charge images of stars collected in time delay integration (TDI) mode with a scanning (rotating) telescope. We focus on the ESA astrometric space mission Gaia, but the results are valid for other scanning telescopes too. The physical attitude of the telescope is shown to be unobservable. Instead, an effective astrometric attitude is observed which represents an average over the TDI exposure time. The effective astrometric attitudes "seen" by different instruments (in case of Gaia: Astro, Spectro, Astro with gates) differ in a non- trivial way. If e.g. the high-precision Astro attitude would be used for the astrometric exploitation of the Spectro data, the Spectro CCDs would be "seen" to float around on the focal plane by several milli-arcseconds. In addition we find that the TDI mode produces an attitude jitter with the period of TDI clocking. We prove that this is negligibly small in the case of Gaia. We point out that the effective instant of observation is not the instant of charge read-out from the CCDs, but about half an exposure time (i.e. up to a few seconds) earlier. This is particularly important for the astrometry of solar-system objects and for the photometry of rapidly varying objects. It is also relevant for all other objects because of the time dependence of aberration. It is not clear whether the differences between the astrometric attitudes of different instruments require separate attitude reconstructions, but an approximate transformation from Astro to Spectro probably will be suffcient.