search

EoRImaging / FHD

Fast Holographic Deconvolution
BSD 2-Clause "Simplified" License
20 stars 10 forks source link

limiting ps_kspan results in a frequency dependent edge #124

Open bhazelton opened 6 years ago

bhazelton commented 6 years ago

This is because if a visibility falls outside the span at any frequency it does not get gridded for any frequency. This causes a bad frequency dependent edge problem. Instead, visibilities should be gridded if they fit inside the span at the gridding frequency.

miguelfmorales commented 6 years ago

Is this actually a bug? The worry was having baselines appear/disappear as a function of baseline. I think both approaches have issues, and I'm not sure which is worse. Just food for thought.

bhazelton commented 6 years ago

I think I'm personally worried more about the frequency-dependent edge interacting badly with the FT, but I can be convinced. Why are you worried about baselines appearing as a function of frequency?

miguelfmorales commented 6 years ago

At least in the frequency direction, a disappearing baseline is very harsh, particularly with the gridding kernel.