The command works as follows: it starts examining the first object in the catalogue table. If the AR of this object does not intercept any other AR, the program continues to establish the window for the fit. Contrarily, if the AR does intercept the AR of another star the command examines whether a third intersection exists and so forth. When no more intersections exist, the program goes on to define the associate subframe to fit.
Going to faint photometric limits the intersections grow because of the higher number of stars and because the AR are larger. The maximum number per window technically acceptable to ROMAFOT is 36, but this figure is normally kept lower (typically lower than 20). The program, in the case of crowded fields, may refuse to group certain objects. This situation will be discussed later.
To establish the subframe to fit, a balance of opposite requirements must be achieved. For instance, since the diffraction creates images with ``wide'' wings one is tempted to integrate over ``large'' windows, causing the undesirable inclusion of many objects. On the other hand, considering that the central pixels of the image are those with higher signal to noise ratios, ``small'' windows could be preferable. However, this necessitates the sky background to be known ``a priori'', a heavy requirement indeed in case of crowded fields!
These considerations and the idea that the sky background should be computed together with the star because the two data are naturally coupled, led to the choice of window-sizes as wide as 9 times the FWHM in the case of isolated objects. If the command is faced with a multiple configuration, the window is determined by a frame, 3 times the FWHM width, surrounding the rectangulus circumscribing the Action Radii.
Often new Action Radii, not intersecting the one under examination, fall into the window. These will be ignored during the fit. To visualise this, a ``hole'', as wide as the relative AR, will be created at the positions of these objects. In this operation some pixels are lost for the fitting, this is the reason for the quite conservative choice of the window size.
After GROUP/ROMAFOT has finished, a histogram of the result (groups, objects which failed to group and so on) is prepared printed in the user terminal and in the logfile.
With the default values the user groups a percentage of all the objects in the list. This fraction varies a lot and depends on the crowding, on the seeing, and on the AR. The latter depends again on the photometric limit requested. Typically, somewhere between 70% and 100% of the catalogue list will turn out to be grouped by GROUP/ROMAFOT.
In case of the default values for AR the command starts to group the objects holding their original AR. Thereafter, if a group exceeds the maximum number of objects (e.g. 15) GROUP/ROMAFOT tries to prepare an acceptable window . This limits the intersections and the groups can turn out to have an acceptable number of components. It is important to note that the size of holes is not affected by this reduction and its original value is conserved.
To group remaining objects, one can execute the command once more with the same catalogue and intermediate table, but with an increased maximum number of objects per window and a reduced AR (in the sense just explained). Here, normally AR reductions exceeding 70% of the original value will produce windows with too many holes, and the final window to fit could contain too few pixels. In this case it is wise to assign a value to the parameter [wnd_perc] in order to provide the fit with enough pixels to compute the sky background.
A situation where a large fraction of the objects are not grouped, even after the AR has been reduced to 70%, can be caused by the following.
Following these considerations one should still be aware that there are advantages with interactive processing. If, for instance, GROUP/ROMAFOT was successful in grouping 95% of the program objects, it may be worthwhile to look at the remaining 5%. This operation to obtain (as a first approximation) complete photometry can take 10 to 20 minutes. Of course, if one is not interested in completeness, these last objects can be dropped, since their photometry will be poor in comparison with the others.
After having grouped the objects, the user tries to fit the subframes by executing the command FIT/ROMAFOT.