search

askap-craco / CELEBI

The CRAFT Effortless Localisation and Enhanced Burst Inspection Pipeline
MIT License
4 stars 2 forks source link

mergeOversampledDiFX.py AssertionError: ncurr < target_nchan #471

Open d-r-scott opened 1 year ago

d-r-scott commented 1 year ago

Encountered for FRB201124a.3

  Traceback (most recent call last):
    File "/fred/oz002/askap/craft/craco/difx/sites/ASKAP/mergeOversampledDiFX.py", line 788, in <module>
      stitchVisibilityfile(difxf,cfg)
    File "/fred/oz002/askap/craft/craco/difx/sites/ASKAP/mergeOversampledDiFX.py", line 557, in stitchVisibilityfile
      assert(ncurr < target_nchan)
  AssertionError

Immediately preceded by a warning:

Warning: discarded an incomplete stitch: ak07-ak15/263/0:1227.5000000/XX mjd:59308.42315615 only 432 of 432 channels

This is the first stitch with ak15 in this timestamp - missing data/bad antenna?

d-r-scott commented 1 year ago

None of the voltage files are shorter than expected - running without ak15 and seeing what happens

d-r-scott commented 1 year ago

Dropped ak15 and then same error happens for ak07-ak18 baseline. Dropped ak07 instead and it happens for ak08-ak15. Maybe it's more to do with the time than the frb? This is only happening for the polcal so I'll keep going without it for now

d-r-scott commented 1 year ago

Nope, happens for the FRB too

adamdeller commented 1 year ago

can you point to the working directory?

d-r-scott commented 1 year ago

@adamdeller /fred/oz002/askap/craft/craco/processing/work/201124a.3/7a/05e9b88854dc54afaf73a567021c10

adamdeller commented 1 year ago

Okay something weird is going on here. In the .difx/ file, there is an extra set of visibilities for the first integration of baselines to the second antenna. This "bonus" visibility seems to have near-zero weight and zero amplitude, and I'm wondering if it has been incorrectly added by padDiFX.

adamdeller commented 1 year ago

Ah yes I see there is an _old.difx, which is presumably from before padDiFX, and it doesn't have this ghost visibility. I think I need to look at padDiFX - I think it must be doing something bad!

adamdeller commented 1 year ago

Hi Danica, the c2_f0 correlation and the c1_f0 reference correlation don't match: the c1_f0 has 20 antennas, while the c2_f0 one has 21! This is why fillDiFX is falling over. Can you check why this might be the case?

d-r-scott commented 1 year ago

This doesn't seem to be an error with the data itself, the data has 21 antennas worth of c1_f0 vcraft files