[daip] edge channel drops

[Eric Greisen]

Lynn D. Matthews writes:
 > I have computed a bandpass calibration for some VLBA data with BPASS using
 > autocorrelation data.
 > 
 > There are solutions in the BP table for all channels, but when I apply the
 > BP table to the data, two edge channels on both sides of the band
 > "disappear"---i.e., they now have zero visibilities.
 > 
 > Does this have to do with the weights assigned to these channels in the BP
 > table/some threshold used by DOBAND? Even though these edge channels
 > are of little use scientifically, on a practical level it is a nuisance
 > to have image cubes with channel numbering different from the uv
 > data---plus, I'd just like to understand what's happening with the data.
 > 

No it is not the weights.  The way that the correlator works means
that the BP tables have to be shifted in frequency a little as a
function of time, source and antenna position.  This causes the looss
of the outermost channels.

These channels are very unlikely to be usable in any case - they are
affected by aliasing in the FFTs.

Eric Greisen