Closed esheldon closed 3 years ago
I'm currently setting the observation ormask to the exposure mask, and leaving bmask as zeros
I don't think this is actually used downstream, only the exp.mask is used, but might want to be consistent with DES here.
@beckermr what do you recommend?
My recommendation is the following:
Then the mask in the exposure would not map to either the obs.bmask or obs.ormask which could be confusing
Indeed it could be, but I think these masks are the most sensible. The OR mask has data about the inputs to the coadd. The bmask has data about how the coadd was processed in each pixel.
now I'm not sure why we want 2 separate from 1 since it is a subset
That's true. We could store them in the same set of bits.
How about this:
Do 1. as you said for ormask, but make bmask a copy of the mask in the exposure
Yep. We should keep ormask flags for which pixels got inteprolated too.
Do you mean define a new mask plane to propagate?
We could re-use INTERP
but then we would not be able to distinguish between things we interpolated and the stack interpolated, which may be useful especially since with real images both may be happening
But this would in any case be redundant with the ormask, which is propagating whatever we would have labeled as interpolated
is this ok to merge?
Did you want to add the CATSIM download the CI job? Happy to do that in another PR too.
thanks, let's do it in a different PR