Bill Thompson & Dave Pike
April 2000
NEW: Sept 2000 Please check out the routine gt_nis_alignment which returns the best estimate of the offset for a particular qlds.
The accurate alignment of the spatial coordinate in NIS1 and NIS2 has always been a concern since it is crucial when comparing properties of small scale features in different lines and attempting to use the plasma diagnostic properties of CDS. Some users have commented, ever since launch, that if some assumptions are made about the structure of spatial features in different lines, then there did seem to be a small offset (possibly at the 1-2 pixel level) between the two wavebands. The extraction windows on the CCD are set by reference to the edges of the slit, which are easily detected, and so any image shift would have to result from optical effects ahead of the slit. However, analysis of existing data never conclusively detected such a shift.
A few months back, George Simnett queried some flare data he had, but since he was comparing different ions it was not clear that the origin couldn't be solar. Recently, however, Terry Kucera and Oddbjorn Engvold showed Bill some data which fairly convincingly demonstrated that there is something more complicated going on beyond simply matching up the data windows to the spectrum edges. Their data appeared to show a ~7 pixel vertical shift between NIS-1 and NIS-2 images.
As a result of that we have looked into the phenomenon again and the results are illustrated in the following plots. The most convenient pair of lines to use are MgIX (368 Å) and MgX (625 Å) since the small temperature difference should minimise any real intrinsic spatial differences in the solar images and these lines have been used throughout the investigation.
Fortunately, since recovery the MgX line has been included in the daily synoptic study and so the synoptic data were analysed by cross-correlating the MgIX and MgX images obtained on 54 days earlier this year. The results for the offset of the two images in pixels are shown below where the results have been averaged for each raster in the synoptic sequence (raster 0 is taken at the north pole and raster 8 at the south pole). The one sigma error bars are also shown. This shows a well defined offset which varies as a function of Solar Y (the only variable in the synoptic sequence!).
Analysis of a more general sample of images resulted in the following plots. In addition to confirming the results from the synoptic data, there is some evidence of a variation of offset with X location.
Such offsets can only be induced by the optics ahead of the slit and since the only way those optics can be sensitive to the pointing is via the OPS, it appears that the movement of the OPS is introducing a distortion into the telescope (the two wavebands use light reflected from different parts of the primary mirror). Analysis of the data shown above hints that the offset is more sensitive to OPS Right leg movement than to the left leg.
While the effect is now obvious (especially with hindsight) for the more northerly pointings it is interesting to revisit pre-loss data. A plot derived from all suitable 1997 data (448 files) is shown below. The Solar X/Y values have been averaged in 100" bins. It appears that the same effect was indeed present pre-loss but at a much smaller level.
For comparison with the synoptic data the pre-loss data only from the same annual time period (Jan-Mar) is shown below.
For the time being users are advised to take careful note of the above results, especially if they are comparing structures with lines in both wavebands. In the meantime we will attempt to determine the offset values more accurately and to confirm the effect's repeatability with a view to providing corrections in the analysis software.
NEW: Please check out the routine gt_nis_alignment which returns the best estimate of the offset for a particular qlds.