The optical spectrum of Procyon was observed from McDonald Observatory
on January 30 and 31 1999. We made use of the Harlan J. Smith 2.7m
telescope and the 2dcoudé spectrograph (Tull et al. 1995) at the
focal station F1, which provided a resolving power of 200,000. Five
different spectral setups (numbered 1 to 5 in the figure below, where
their location in the spectrograph's focal plane is shown) provided almost
complete spectral coverage from 4559 to 5780 Å, and different
exposures (up to ten for a single spectral range) were coadded to
reach signal-to-noise ratios in the range 550-2000.
ACCESS THE DATA
The absolute wavelength scale of the atlas has not been corrected for the different systematic shifts that affected the observations (orbital motion, systemic velocity, 'heliocentric' correction, etc.). The relative shifts between different spectral setups were corrected as described in Allende Prieto et al. (2002).
These cross-dispersed echelle spectra were continuum corrected
by fitting a smooth function to the different orders. These removes effectively
slow instrumental distortion of the flux, such as the blaze function, and
works reasonably well in spectral regions populated by weak lines, but
it miserably fails in the proximity of strong lines, whose widths approach
or exceed the spectral coverage of a single order. A more adequate treatment
for regions around strong lines has been developed and described by Barklem
et al. (2002). Careful application of these techniques to these data
in the orders close to HBeta gives much better results in this region.
If you worry about the continuum scale in the HBeta region, the following
data should, after correcting for a systematic velocity offset, replace
the corresponding segment in the atlas above:
Because the atlas is made from the combination of different orders, systematic errors are likely to happen at the joints. Small discontinuities in the flux will signal such places. If you cannot see some of them, that means the overlapping fluxes from adjacent orders matched very well and systematic errors are then small.