With the linearized fluxes extracted, the wavelength system of each spectrograph camera could be calibrated by means of a reference calibration spectrum by remapping the (x,y) positions of its emission lines from the original geometry to the rectilinear detector surface. For this step one may think of the dispersion axis as ``x" and of the spatial (or echelle order m) axis as ``y." Wavelengths of the emission lines were taken from the study of Reader et al. (1990), which was commissioned for the purpose of calibrating wavelengths of the HST spectrographs. The first of four steps in the NEWSIPS wavelength calibration procedure was interactive and consisted of running the IRAF routine ecidentify to map the positions of a set of lines to wavelengths for a reference WAVECAL echellogram. In this step a quadratic Cheybshev polynomial functon was used to map the (x,y) pixel locations in the reference echellogram to a wavelength table. The second step was to utilize the IRAF routine ecreidentify Chebyshev solution determined for the reference echellogram in to determine predicted pixel positions for a few hundred lines in all the WAVECAL observations for a given camera. A polynomial fit was performed for each of the echellograms, resulting in a set of wavelength zero-points and mean pixel-to-pixel wavelength increments for each order. In practice, the wavelength zero-points for each ``science echellogram" were fit to a cubic polynomial in both time and camera-head temperature (``THDA"). These fits were adopted for all science data but of course not for the WAVECAL calibration observations themselves. This last step compensated for image shifts statistically by means of a previously derived least squares correlation of wavelength shift as a function of time and temperature. Note that the wavelength calibration of the WAVECALs technically applies to observations made through the small aperture. However, this calibration can be readily applied to large-aperture observations from prior measurements of the offset between the two apertures in the dispersion direction.
We should pause to point out that whereas the wavelength calibration for NEWSIPS was performed with a simultaneous polynomial fit in both the wavelength and echelle order dimensions, the number of WAVECAL emission lines was sufficient to permit a calibration of wavelengths for most echelle orders separately (Smith 1991). The first (``global") approach is clearly the preferred one if pixel-to-pixel spacings on the detector are regular because a large number of lines can contribute to the solution for any individual order. However, during the IUE lifetime the cameras suffered both electro-optical distortions and differential shifts arising from shears between fiber-optical bundles within the coupling-plate of the camera. A global wavelength solution tends to smooth over these small-scale distortions, causing correlated order-to-order errors in derived wavelengths. Consequently, with exceptions noted by Garhart et al. (1997), the ecidentify steps described above were implemented for individual spectral orders in the NEWSIPS calibration.