VOSA adds a sistematic error of 0.3 mas to the original error, as recommended in Brown et al. 2016.

We always take the nearest counterpart within the search radius chosen in the "VO Photometry" tag. For those catalogues containing both point and extended sources (e.g. SDSS, UKIDSS, VISTA, DES,...), if the nearest counterpart is an extended object, then VOSA does not return any photometric information.

In summary, points with ΔFlux=0 are treated as if they had the largest error in the SED. VOSA calulates the largest relative error in the SED, adds a 10% and then assigns this relative error to those points without an observational ΔFlux. See the Fit help section for details.

For pogson magnitudes, being ${\rm F}_0$ the photometric system zero point flux: $${\rm mag} \pm \Delta {\rm mag} \Rightarrow {\rm Flx} \pm \Delta {\rm Flx}$$ $${\rm Flx} = {\rm F}_0 \ 10^{-{\rm mag}/2.5} $$ $$\Delta {\rm Flx} = {\rm Flx} \cdot \Delta {\rm mag} \cdot \ln(10)/2.5$$

No. They are different catalogues. The number of sources in common is less than 50% and Hauck et al. (1997) has even more sources than Paunzen (2015).

There are two main reasons for this effect:

• Sometimes, the magnitude given by the catalogue is just the average of different measurements taken at different epochs by different groups. In this case the error in the magnitude is the standard deviation of these measurements which may be large in some occasions.
• Sometimes, magnitudes in the Paunzen catalogue have no associated errors. In this case, VOSA assigns to these magnitudes, in the chi2 process, the largest of the photometric errors multiplied by 1.1. If you visualize the SED in the chi2-fit tab, you will see these large asigned errors.

The catalog provides: $$ V \pm \Delta V $$ $$ (b-y) \pm \Delta (b-y) $$ $$ m1 \pm \Delta m1 $$ $$ c1 \pm \Delta c1 $$ and we calculate: $$ y = V $$ $$ b = (b-y) + y $$ $$ v = m1 + 2(b-y) + y $$ $$ u = c1 + 2m1 + 3(b-y) + y $$ $$ \Delta y = \Delta V $$ $$ \Delta b = \sqrt{ \Delta (b-y)^2 + \Delta y^2 } $$ $$ \Delta v = \sqrt{ \Delta m1^2 + 4\Delta (b-y)^2 + \Delta y^2 } $$ $$ \Delta u = \sqrt{ \Delta c1^2 + 4 \Delta m1^2 + 9\Delta (b-y)^2 + \Delta y^2 } $$

References of all used catalogues can be downloaded from the "Download Results" tab. In the case of photometric catalogues in which a counterpart to the target exists, They will be included in the "info/refs.dat", "info/refs.bibtex.bib" files, regardless the utilization of these points to build the SED.

On the contrary, if the target has no counterpart in a catalogue, this catalogue will not be included in those files.

The fact is that Gaia G filter is very wide compared to most filters with similar wavelegths. Thus, it averages the spectrum over a large wavelenth range. If it happens that the spectrum is steppy in that range, the photometric point will typically lie far from the spectrum and probably outside the main SED tendency.

This is just an example of the fact that you shouldn't try to fit observed photometry using the theoretical spectrum directly. You need to compare the observed photometry with the synthetic one calculated using the theoretical spectra and the filter passband.

Magnitudes should not produce negative fluxes but SDSS magnitudes are not the typical pogson ones but asinh "laptitudes" and the conversion formula that we apply is: $${\rm Flx} = {\rm F}_0 \ 10^{-{\rm mag}/2.5} [ 1-{\rm b}^2 * 10^{2 {\rm mag}/2.5}]$$

this shouldn't produce negative fluxes either but it can happen and, when it happens, VOSA rejects the corresponding flux values as bad.

You can take a look to the filter information for more details and the particular parameter values for each SDSS filter.

If the precise value of Av is not known and you choose to include it as a fit parameter, and specially when no photometric information in the blue range is available, it may happen that different combinations of extincion and effective temperature give very similar fits leading to a Av/Teff degeneracy.

This effect is considerably reduced when the distance to the object is known so that you can restrict to small values of Av.

This can be clearly seen if the Bayes analysis is performed (see figures). In this case, the best effective temperature calculated using the chi2 fit may not be the good one from the physical point of view.

Some theoretical spectra like, for instance, the BT-Settl-CIFIST ones, are very big files. If you decide to include the spectra in the plots, VOSA, during the processing of the fit results, has to ask the corresponding VO service for a degradated version of the best fit spectrum for each object. The VO service makes the calculation to degradate the spectrum, returns it to VOSA and VOSA includes it in the plot. This operation takes longer for big spectra (like BT-Settl-CIFIST), maybe a few seconds, but the acumulated time excess for many objects can be very relevant.

Thus, if you have a VOSA file with thousands of objects, don't check the "Include model spectrum in fit plots?" unless you really need that.

The theoretical spectra are not included in the list fo VOSA products mainly because of their size. For instance, the size of a single BT-Settl spectrum is 8 MB.

The best way to download the theoretical spectrum that best fits the data is going to the "Best fit" table of results. The last column of the table, titled "Data VOTables" gives you a link to get the full theoretical spectrum corresponding to each object fit.

WARNING: As these files may be large, and in order to avoid web browser crashes, it is advisable to save them using the "Save as" option (right buttom of the mouse) instead of directly cliking on the link.

Besides this, as a side trick, take into account that .agr files for the plots with the fit results contain a resampled version (lower resolution) of the theoretical spectrum if you have chosen to include spectra in the plots. These are ascii files, and you can find the spectrum at the end of the file as the table with the largest number of points.