1 Introduction

Satellites I - V of Uranus have orbits closely aligned to the planet's equatorial plane. However unlike any other planet in the Solar System the equator of Uranus has an inclination near 90$^\circ$whereas the rest are closer to zero. The satellite orbits were widest open in 1986 and are edge-on 21 years before and after. The eccentricities are best determined while the orbits are open and the inclinations when seen edge-on.

Although the surfaces of Uranus and his satellites can be studied from space, their interior structure can be studied only from their dynamical interaction. In principle their mutual perturbations should reveal their masses and the dynamical compression of the planet J₂; but Taylor (1998) has shown that it is not possible to determine the compression and the masses of the five largest satellites simultaneously. This indeterminacy is not fundamental but arises from the limitations of the dataset, already confined to only the best modern observations. The objective of the present investigation is to add to the dataset in both size and accuracy with a view to removing this indeterminacy. A preliminary reduction of these observations was used by Taylor (1998); those presented here have been reduced with the more elaborate procedure described below.

At the 1990 and 1991 oppositions we observed Uranus with his satellites through a Gunn Z filter (effective wavelength $\sim 0.9\ \mu$) with a GEC CCD. The field of the chip was $3\hbox{$^\prime$}\times 2\hbox{$^\prime$}$. Contemporaneous observations of the satellites of Saturn have been published by Harper et al. (1997) with the same instrumental set-up. Harper et al. had insufficient data to calibrate the scale and orientation of the chip independently, and so resorted to calibrating with the well-known orbits of the brighter satellites.