This is a continuing frustrating story. The latest version of qac_bench5 , which contains the M100 re-gridded data, has the first or last channel masked, depending if you use SDINT or Feather.
In earlier versions the first channel would have not all data included, resulting in a different beam, and forcing us to clean with a beam_per_channel.
Unclear if this is a python rounding, or casa meta data (e.g. restfreq) issue.
This is a continuing frustrating story. The latest version of qac_bench5 , which contains the M100 re-gridded data, has the first or last channel masked, depending if you use SDINT or Feather.
In earlier versions the first channel would have not all data included, resulting in a different beam, and forcing us to clean with a beam_per_channel.
Unclear if this is a python rounding, or casa meta data (e.g. restfreq) issue.