Unlike JWST, IRIS will be able to concurrently readout the subarrays (up to 10 rectangular, any location, any size, not overlapping), and the rest of the frame.
Therefore we will have to process frames which are full except those rectangular holes, and then somehow combine them back properly into a single frame.
Plan is to have a FITS extension, 16 bits, so 1-10 will be identifying which pixel is part of which subarray. It is 32MB for this, compared to 64MB of data, however this is only when we are running with subarrays, so should be fine.
I also think we should have one bit of the standard DQ mask which indicates that a pixel is part of any subarray, so that it is handled exactly like dead pixels and so on. So the pipeline should be already able to just process it (will test it).
Unlike JWST, IRIS will be able to concurrently readout the subarrays (up to 10 rectangular, any location, any size, not overlapping), and the rest of the frame.
Therefore we will have to process frames which are full except those rectangular holes, and then somehow combine them back properly into a single frame.
Plan is to have a FITS extension, 16 bits, so 1-10 will be identifying which pixel is part of which subarray. It is 32MB for this, compared to 64MB of data, however this is only when we are running with subarrays, so should be fine.
I also think we should have one bit of the standard
DQ
mask which indicates that a pixel is part of any subarray, so that it is handled exactly like dead pixels and so on. So the pipeline should be already able to just process it (will test it).@arunsurya77 @ikashell any comment?