The data extracted from the NIS detector can be compressed prior to being downlinked. Several options are available and the choice forms an integral part of the raster definition. Compression is not considered necessary for GIS data owing to its paucity.
The NIS options generally used are:
1) 12-bit truncate. The mismatch between the 12-bit CCD data and the 16-bit telemetry normally means 4 redundant bits in the telemetry. In this mode consecutive 12-bit data values are packed into the 16-bit telemetry to avoid empty telemetry bits.
2) Sum over total data window, or over wavelength within a window. Drastic compressions, but which nevertheless have their uses.
3) Sum over N rows within the data window. Since the spatial resolution of CDS is less than the 1.68" of a single CCD pixel, little information is lost by summing at least 2 rows in the N/S spatial dimension. Beware that although this leads to a factor of N reduction in the raw data, no bit packing is then possible and furthermore the extra on board processing takes a not insignificant percentage of the expected time saving. The only way to check the effect on the telemetry and hence the raster duration is to test run the raster on the EM at RAL.
All compression of the data is reversed before the data are written to the FITS data files.