As discussed originally in JP-3250, NaNs in the flat variance array were previously contaminating the ERR estimates of extracted 1d spectra. Most of the reasons for that have now been solved via that and other linked tickets. A final aid will be provided by JP-3570 which will ensure consistency where SCI=NaN, ERR=NaN, and DQ=DO_NOT_USE prior to resampling/cube building.
As such, there should now never be pixels that have finite SCI values but NaN ERR values, etc.
However, as a safety precaution this ticket is to ensure that any such NaN values in the ERR array are ignored during spectral extraction, as they should have DO_NOT_USE set anyway.
stscijgbot-jp
commented
2 months ago
Issue JP-3655 was created on JIRA by David Law:
As discussed originally in JP-3250, NaNs in the flat variance array were previously contaminating the ERR estimates of extracted 1d spectra. Most of the reasons for that have now been solved via that and other linked tickets. A final aid will be provided by JP-3570 which will ensure consistency where SCI=NaN, ERR=NaN, and DQ=DO_NOT_USE prior to resampling/cube building.
As such, there should now never be pixels that have finite SCI values but NaN ERR values, etc.
However, as a safety precaution this ticket is to ensure that any such NaN values in the ERR array are ignored during spectral extraction, as they should have DO_NOT_USE set anyway.