elpajare I am more concerned with the magnitude than the number of stars. It would be interesting to know what the limiting magnitude is for Asiair to take them into account.
Like you, I do not have source code to look at. Its a shame, because many of the outstanding bugs would be found in a few hours by cloud sleuths.
However, from casual observations, the Detect Star tools and the Plate solve tools appear to use different number of stars. It makes sense because the former cannot use saturated stars to estimate the HFD (notice it leaves out saturated stars), while the latter has to admit every star to create the asterisms (check out astrometry.net's document; ASIAIR has to be using something similar).
Brighter star magnitudes are important of course, due to better signal to noise ratio. But there are other factors too, sensor read noise, sensor gain vs exposure time, sky background (SQM or Bortle), filters, tracking stability,...
I also suspect that it is better to have many stars with similar magnitudes, instead of a couple of very bright stars, and the rest too dim to use.
I don't know how ZWO picks stars to include in the ASIAIR plate solving database. If it were me, I will not pick a fixed magnitude to include in the database, but pick stars so that the star density (or more accurately, the density of asterisms) remains moderately constant (dimmer threshold in star-poor areas of the sky). This will keep the database small, while being less finicky with camera angles and where in the sky you are plate solving, and it may give a smaller FOV on average (why include as many stars in the middle of Cygnus, for example).
Chen
