In this study we cross-correlate many IUE echellograms of a variety of well observed stars to evaluate systematic error sources in the wavelength zero-points ( velocities ) of all three cameras .
We first evaluate differences between the final archived ( “ NEWSIPS ” ) and the originally processed ( “ IUESIPS ” ) echellograms .
These show a marked time dependence in zero-point for the SWP camera due to several revisions of wavelength calibration coefficients used for IUESIPS .
Smaller offsets are present for the LWR camera between the two processings .
We also evaluated small-amplitude fluctuations in the zero-points of the NEWSIPS wavelength calibration spectra themselves .
In the case of the SWP camera , these variations are too complicated to have been completely removed in the NEWSIPS wavelength calibration .
We also examine wavelength zero-point disparities between data obtained both through the small and large entrance apertures as well as for observations made by different target acquisitions of faint and bright stars .
We also find that statistical differences between these alternative observing modes are virtually nil .
For large-aperture observations the dominant error source is the placement of the target in the aperture .
These give rise to non-gaussian , extended “ tails ” in apparent velocity .
We also searched for spurious trends with time .
Except for a possible trend for faint objects with SWP camera data , we can not detect significant dependences with time .
Additionally , we discovered a trend with telescope focus for datasets derived from intensive monitoring campaigns of bright stars .
These exhibit a repeatable , one-day “ radial velocity variation ” with a semi-amplitude of nearly 3 km s ^ { -1 } .
This pattern appears to be a by-product of fluctuations in telescope focus caused by operational procedures to maintain the ambient instrument temperature .

In the second part of the paper , we measure the mean zero-point errors of NEWSIPS echellogram data with respect to laboratory results by using the Goddard High Resolution Spectrograph ( GHRS ) spectral atlas of the O9 V spectral standard 10 Lacertae as an intermediary reference .
We find that the derived apparent velocity difference for this star is essentially zero : -1 { \pm } 3.5 km s ^ { -1 } .
Several less precise comparisons lead to similar results , including cross-correlations of : ( 1 ) spectra of 10 Lac and two stars with similar spectra , HD 93521 and HD 60753 , ( 2 ) lines in common with the SWP camera and GHRS and STIS atlases of Arcturus and Procyon , and ( 3 ) interstellar lines in the GHRS spectrum of the white dwarf G191-B2B .
The zero-points of the NEWSIPS -processed long-wavelength cameras are evaluated and are also found to be nearly zero ( { \pm } 5 km s ^ { -1 } ) relative to the Arcturus and Procyon atlas calibrations and relative to one another .
In general , these results do not support the suggestion by González-Riestra that corrections should be introduced to the wavelength scales of various NEWSIPS high-dispersion data products .
Despite our optimistic results , it is obvious that using small IUE datasets from large-aperture observations of arbitrarily chosen stars can contain velocity errors of at least a few km s ^ { -1 } .