The determination of an accurate dark+stray light signal for the L04 observations was more difficult than for the L03 dataset. As the pointing towards the source was different, the L03 dark values could not be applied to the L04 observations. Also they could not be determined from L04 non-prime observations because there was insufficient wavelength coverage to recover accurate values. This meant that the dark signal for each L04 observation had to be estimated by comparison of the continuum level with an LWS grating observation. A grating observation with similar pointing to the L04 observations was used (TDT 46400310). It was assumed that the continuum was well calibrated in this observation. This is reasonable for the short wavelength detectors as they were not affected by the detector non-linearities that caused large uncertainties at longer wavelengths. The dark signal for the L04 observations was adjusted until the continuum matched with the grating observation. At this stage one of the L04 observations (TDT 32601007) was rejected from the analysis as a dark signal could not be found that caused both the line depth and continuum level to be consistent with the other observations.
L04 reduction with FP_PROC was slightly different to that for the L03 observations as there was generally only one L04 mini-scan. This meant that it was more difficult to determine the best shift to use with the interactive grating shifting routine. In order to make this easier, the grating L01 measurement and the L03 prime data were over plotted whilst making the shifts. It was important to determine the correct position for the grating profile centre so that the L04 observations could be co-added. One observation (TDT 46201118) was rejected from the analysis at this stage because it did not have full coverage of the line. This meant that the continuum level was only reached on one side of the line and mini-scan shifting was highly uncertain.
This left three good L04 observations that covered the line and adjacent continuum. These were deglitched using the same method described for the L03 observations. Deglitching was particularly important for these observations as there were 40 repeated scans. Large glitches occurring across the line in several of these scans caused a significant difference to the line depth before they were removed. Each repeated scan was examined individually and compared with its neighbours to remove all glitch signatures. The three observations were then co-added to produce a spectrum with 120 repeated scans.