3. Important points of the data processing
3.1. General management
The practical implementation of the processing is briefly described in Paper I, Sect. 4.2. There is a double iterative process: the first one consists in improving the reference values at the step i by injecting the solutions obtained at step ; the second one corresponds to the iterative filtering of the largest residuals found in this step. The data of the 145 selected objects have first been processed by an all purpose algorithm, quite similar to the one used for the simulation. However the poor overall quality of the results has shown that a specific processing adapted for each star was preferable. One of the most crucial points to ensure the convergence of the processing was the choice of the threshold to filter out the residuals and this had to be fine-tuned on a case by case basis alongside the weighing of the observations.
3.2. The reference values of and B
As it was mentioned in Paper I, Sect. 4.2, the equations relating the Hipparcos observations on the circle to the astrometric and physical parameters are all non-linear. Thus, it is useful to start the processing with input values of and B relatively good. It turns out that for all the cases presented in this paper, the magnitude difference is fairly well known and this was sufficient to compute an input . On the other hand, for more than half of the systems, no reliable input value of the mass ratio B has been found (see Table 9). In this case the mass luminosity relation for dwarf stars has been used to yield an approximate first guess of B via the expression,
where and are the masses of the components.
3.3. The choice of an orbit
For many of the systems, there are more than one set of orbital elements proposed in the literature. We have tested systematically all the possible orbits for each object and then selected a particular set of orbital elements. The selection criteria were based on the quality of the fit (Unit Weight Variance), the number of required iterations and the stability of the solutions to a small perturbation of the input values and of the orbital elements. For most of the cases, the selected orbit (see Tables 4 - 5) happens to be the most recent one. It must also be noted that the elements derived from the speckle interferometry were particularly satisfactory, regarding their adequation to the Hipparcos data.
© European Southern Observatory (ESO) 1998
Online publication: January 16, 1998