In figure 5 we show a rebinned aar with the new
calibration divided by the OLP10 AAR. This can be compared to the
mean ratio curve of the STARTYPE group (figure 1). We
see a qualitative
agreement in the differences around 4.5 
m and the difference in
the
short-wavelength end of 2C. We interpret the 2% features in the 9-12
m region in the curve in Figure 1 as noise in the
residues due to large differences in memory effects.
During the scientific validation of OLP 10.0, L. Decin described two wavelength regions where the RSRF correction could be improved in band 2A and 2C. The features she described in the comparison of synthetic spectra and OLP10 SWS spectra compare well with the changes to the RSRF calibration for OLP10.1.
We believe the accuracy of the absolute and relative flux calibration has improved with the new delivery. The improvement, however, is at a level smaller than the typical reliability of individual band 2 observations because of memory effects. An example can be seen in Figure 6. We have plotted an AOT1 and an AOT6 observation of sirius, processed with OLP 10 and the new calibration. One can see that the difference between the two observations (with different scanning strategies and hence different memory effect regimes) are comparable to the differences between the two calibrations (we only consider the short wavelengths, at longer wavelengths the situation only gets worse).
 |
We should be aware of the fact that, although we might have taken away concerns about biasing to errors in synthetic spectra, there is a risk that the (1) low number of observations, (2) the similarity of the spectral shape in the observations and (3) the nearly identical observation mode (AOT1 speed4, Vega is observed with speed 3) in the new calibration can introduce a systematic bias to the typical memory effect signature in AOT1-speed 4 observations of hot stars. We have verified that this bias is in any case not worse than the bias previously present in the olp 10 calibration.
