This article describes the wavelength calibration phase of data processing. The FIMS and SPEAR teams both followed the same procedure for this step, except for the final trim of vignetted edge wavelengths performed by the FIMS team.
The preceding steps were applied to the aurora and sky survey data to provide spectra with features suitable for wavelength calibration (See Figure 1 below). For the long and short channels, 20 and 13 identifiable lines are found, respectively. Apart from the HI Lyman series, most of the observed features were blends and their effective wavelengths were estimated using the Atmospheric Ultraviolet Radiance Integrated Code (AURIC; Strickland et al. 1999). Mean centroids for each line were determined from the entire height of the detector (minus the bright edges). The dispersion is very nearly linear for both channels.
Figures 1 - (a) corrected L-band aurora, (b) corrected S-band aurora, (c) corrected L-band sky survey, (d) corrected L-band aurora.
Second order polynomial fits of wavelength vs pixel showed an RMS error of 0.23 and 0.12 Å for the long and short channel, respectively (Figure 2). Higher order polynomials did not reduce the RMS error significantly. Since the dispersion is so close to linear, extrapolation of the polynomial fits to within a few pixels of the detector edge is relatively robust.
Figure 2 - Plots of residuals from second-order polynomial fits to the dispersion solutions for L-band (upper) and S-band (lower) as a function of wavelength.
However, as in all micro channel plate detectors, the edges show very strong non-linearities in both position and sensitivity and cannot be calibrated reliably.
In products produced by the FIMS team, the edge wavelengths have been masked and removed. In products produced by the SPEAR team, users should trim these vignetted wavelengths themselves.
Note: a slightly different wavelength calibration procedure is described by Lee et al. 2004, but was not implemented in either team's products as currently provided.