In this paper, the ASSI recombining table has been presented. The star tracker is described in details. It
is made of two tip-tilt mirrors (one per telescope) and one quad cell equipped with four PMTs. The
servo-loop of this system is characterized by its transfer functions. The closed loop bandwidth at
-3 dB
is 63 Hz while the rejection bandwidth at 0 dB is 22 Hz. The theoretical behavior of the quad cell
signals is presented, showing the requirement of a loop gain adjustment versus the seeing conditions.
The theoretical behavior of the temporal power spectrum of the turbulence angle of arrival is also
given. We emphasize the knee frequency of this spectrum which allows a rough estimation of the
average wind speed in the atmosphere. While the power of the spectrum is directly related to r0. We
were able to estimate both r0 and wind speed using the angle of arrival data recorded in closed loop
operation, allowing a full characterization of the observation conditions.
Performance of the ASSI star tracker was measured on the sky during a few nights. It allowed to bring
a significant gain in visibility measurements on the I2T, which will be presented in a future paper. The
wavefront angle of arrival correction system on the I2T + ASSI instrument proves to be able to reach
good performance in terms of image tracking accuracy, 
arcsec (
cm at 600 nm). The
residual wavefront angles of arrival cause a reduction in the fringe visibility measurements. The rms
visibility resulting from losses caused by small angle of arrival errors when the beams are combined are
given by Buscher (1988). The arcsec angle of arrival error implies that the two beams are kept to
within one third the size of the Airy disc radius at 
, and ensures visibility loss below 15%.
We observed a permanent deficit in tracking accuracy for the South channel at relatively low light level
which will require to be fixed in a near future.
Most of the measured temporal spectra of angle of arrival exhibit a behavior compatible with the
Kolmogorov theory, in terms of power law.
Estimates of 10 cm for the Fried parameter r0 and 5 ms for the coherence time t0 at 600 nm have been
met in our observations, and are presently considered as average values on the site, yet requiring a better
statistic. Such values are comparable with those obtained on other sites
(Colavita et al. 1987;
Nightingale & Buscher 1991; Davis et al. 1995).
Acknowledgements
This work has been supported by contracts from "Direction des Recherches, Etudes et Techniques (DRET)'' of the Defense Minister, France. We wish to thank C. Coudrain and G. Merlin for their helpful interventions on the instrument.