This plot shows the relation between SNR (called 'mean_s2n' in the database) and average counts (not flux, although it is called 'mean_flux') for the GIRAFFE IDPs. Both parameters measure the quality of the reduction process: the higher the flux, the higher the SNR if the noise is dominated by photon noise. In that case a square-root law is expected. If there are other noise sources (e.g. incomplete removal of gain noise, fringing, etc.) deviations from the square-root curve should be expected, e.g. a plateau in the SNR curve for spectra with high counts, or an offset at very low counts. See the plot tutorial for more.

Note that in the period 2006-2008 these plots indeed show a downward SNR deviation. This was due to master flats generated with an earlier pipeline version that had a bug in the error computation for those products. Since the earlier master flats have been recreated from scratch, with the same pipeline version as the 2008+ ones, the corresponding IDPs do not have that SNR deficiency. It is spurious anyway, the IDPs do not suffer from a real SNR degradation.

We have selected the brightest fibre in each exposure for display. For each data point from the brightest fibre there are on average 50-100 more from fainter fibres. No distinction has been made between Medusa1 and Medusa2 slits. The selection of settings is based on statistics.

Both parameters are averages across the whole spectrum. They depend therefore somewhat on the slope of the spectra and therefore on the nature of the targets. This might explain the observed spread in the relations, which however are all close to square-root laws.

The upper and lower boundaries of the "fan" are scaled square-root laws and purely empirical. They are plotted for orientation and are not fits.

NOTE: don't try the 'Data downloads', they take very long or even time out.