2 The observations, measures and data reduction

The observations were carried out at the Cassegrain focus of two telescopes at the Observatório do Pico dos Dias (OPD) of the Laboratório Nacional de Astrofísica - Brazil. About 70% of these observations were made at the 1.6 m Perkin-Elmer reflector for wich the scale at the focal plane is 13''/mm and 30% at the 0.6 Zeiss Jena reflector for wich the scale is 27$\hbox{$.\!\!^{\prime\prime}$}$5/mm. The latitude of the observatory is $\phi\approx -23^\circ$ and so Neptune's culmination is near the zenith of those years. For details about the telescopes we refer to Veiga et al. (1987).

For all of the 229 observation made at the 0.6 m telescope we used the CCD EEVP8603A which is an array with 385$\times$578 square pixels, each measuring $22\,\mu$m, which corresponds to $0{\hbox{$.\!\!^{\prime\prime}$}}605$. The 530 observations made at the 1.6m telescope used the EEV CCD-05-20-0-202 which has an array of 770$\times$1152 square pixels with $22.5\,\mu$m or $0{\hbox{$.\!\!^{\prime\prime}$}}293$(for 319 observations) and the CCD SITe SI003AB wich is a square with 1024$\times$1024 square pixel with $24\,\mu$m corresponding to $0{\hbox{$.\!\!^{\prime\prime}$}}312$ (for 211 observations). No filter was used for most of the observations. However, for 53 observations made in 1997, an Johnston R filter was used.

The exposure time varied from 1 to 5 seconds depending on the meteorological conditions and on the utilized devices.

We use the program ASTROL (Colas & Serrau 1993) to find the centers of the planet, satellite and stellar images. For the determination of each center we take a small area containing the image and a bi-dimensional Gaussian is fit to this image. The Gaussian is added to a second degree polynomial as to remove the sky background. The errors upon the centering procedure were $0{\hbox{$.\!\!^{\prime\prime}$}}04$ for Neptune and $0{\hbox{$.\!\!^{\prime\prime}$}}02$ for Triton. For a discussion on the determination of astrometric centers see Veiga & Vieira Martins (1995).

For the astrometric calibration we made an adaptation of the method of the secondary catalog to the special conditions of the small CCD fields. This method, which is presented and tested in Vieira Martins et al. (1996) (see also Assafin et al. 1997a,b), consists on the setting of an astrometric catalogue for the stars on the CCD, using their images in the Digitized Sky Survey and the positions of nearby stars from the Guide Star Catalog corrected by the PPM Catalog.

To correct the systematic errors due to the color difference between Neptune and Triton for the frames took without filter, we computed the total astronomical refraction separately for Neptune and Triton following the procedure presented in Paper I. In few words, we considered a difference of $0{\hbox{$.\!\!^{\prime\prime}$}}19$ for the planet and the satellite refraction constants which corresponds to $0{\hbox{$.\!\!^{\prime\prime}$}}15$ after the corrections due to the usual values for the pressure and the temperature during our observations. This value corresponds to a difference of $0.07\,\mu$m between the effective wavelength observed for Neptune and Triton. No correction was made for the observations taken with the R filter.

In Table 1 (accessible in electronic form) we list the observed positions of Triton relative to Neptune. The data are presented in the following form: the first line gives the year, month and day and decimal fractions of UTC days, corresponding to the mean instant of the observation. In the next line we list the name of the satellite followed by $X (\Delta\alpha {\rm cos}\delta)$ and $Y (\Delta\delta)$ in arcseconds, referred to Neptune. The reference system is referred to the equator and equinox of J2000.