There is no doubt that for radial velocity or wavelength studies involving IUE data, the NEWSIPS processing is much superior to that of IUESIPS. The wavelength calibration coefficients were periodically updated during the mission in IUESIPS, so datasets spanning several years cannot have consistent wavelengths. In this section we quantify the zero-point differences between IUESIPS and NEWSIPS calibrations during the mission.
For this purpose we chose hot stars that the IUE observed over nearly
the entire satellite lifetime.
The first two stars, 10 Lac (09 V) and 
Sco
(B0.2 V), are sharp-lined spectral standards with no known radial velocity
variations. We obtained all SWP high-dispersion observations from the
MAST database after 1979.0 (except for SWP 01205, 01765, & 02051 for
which IUESIPS data were not available) and cross-correlated pairs of
IUESIPS and NEWSIPS fluxes in spectrum segments in orders m =
50-119. These results are shown in Figure 4.
The series of dotted vertical lines
mark the dates in which then newly-derived wavelength calibrations
were implemented into IUESIPS processing.
We also cross-correlated 33 echellograms from both processing systems
for six white dwarfs which were analyzed by
Holberg, Barstow, & Sion (1998; hereafter ``HBS").
These stars are WD 0005+511, WD0044-121, WD0621-376, EG-102, GD-394,
and WD2211-495, and in all cases their spectral lines are sparse and broad.
Thus, we found it advantageous to cross-correlate the same 50 orders
of these spectra as those utilized for our study of
10 Lac and Sco. We also cross-correlated a few orders
dominated by interstellar lines.
Figure 4
Cross-correlation shift differences in the sense IUESIPS - NEWSIPS
with time. Displayed are shifts for echellograms observed for the
velocity standards (as labeled). The white dwarfs are those
studied by Holberg et al. (1998). Vertical dotted lines
indicate the times in which new wavelength calibration constants were applied
to the IUESIPS processing software according to Thompson et al. (1981),
Thompson et al. 1983, Thompson & Turnrose, 1983, Gass & Thompson 1985, &
Garhart 1993).
We found a mean zero-point difference, IUESIPS - NEWSIPS,
among the sample of HSB white dwarfs of +9.5
± 1 km s-1.
This compares rather well with the difference of
+9.1 km s-1 given by HSB (referred to 
1400).
The dependences of zero-point with time for both white dwarfs and hot main sequence stars in Fig. 4 are likewise similar to the HBS results. Thus, the form of this zeropoint-difference curve is insensitive to the sample of stars chosen. Rather, it is largely determined by the dates of the revisions of the IUESIPS calibration constants were implemented (see vertical dotted lines in the figure). Notice that the net difference, IUESIPS - NEWSIPS, is positive for all epochs and increases during the late-1980's to a rather well determined maximum of +18 ± 1 km s-1 at 1992.5. It then decreases quickly during 1993 and is perhaps constant thereafter. The IUESIPS wavelength calibration was unsettled prior to 1983, producing a large scatter compared to the homogeneously-calibrated NEWSIPS system.