@rstofi We can chat about the correctness of the output here.
We need to clarify what we determine as correct here. I'll read through the email thread and summarise what's been said:
we want the smeared grid to have the same non-zero value occupancy as the psf and visibility grids (you were unsure of this in the last email so we need to make sure)
we need to check some of the changes to the yandasoft code, so as to ensure that the output is the same after the conversion to python.
The reason for this whole setup is
the psf and visgrids are not corrected for the anti-aliasing (AA) kernel.
the pcf stores the average grid values without accounting for the AA kernel
we need to make one of two corrections:
uniformly add the AA kernel size to the pcf grid values
only add the AA kernel size to the pcf grid values that are less than the AA kernel size
we simply add the values, which may not be correct
these should be computed from the convolution
ie. the convolved kernel size won't be the same as the sum of the two kernels
My knowledge in this subject is limited, so please check my summary and make sure I haven't missed anything or misinterpreted anything.
@rstofi We can chat about the correctness of the output here.
We need to clarify what we determine as correct here. I'll read through the email thread and summarise what's been said:
My knowledge in this subject is limited, so please check my summary and make sure I haven't missed anything or misinterpreted anything.