[daip] BPASS bug for single-source files

[Eric Greisen]

Michael Bietenholz writes:

 > Does that mean that you can't do a proper BPASS solution on a single
 > source file where there is no longer a CL table, or is the freq. in the
 > header of the single-source sufficiently accurate?
 > 
 > > I propose remove this CL table sort entirely and changing BPASS to
 > > tell FRQUPD that the table is neither open or sorted on the first
 > > call, but that it is open thereafter.

     I think we have been doing fine without this last bit of accuracy
- we do not use it for cross-correlation bandpasses since the
frequency is different for each antenna.  The shift is done after the
solution in that case and is probably accurate enough.  The shift is
small most of the time - so small that we had trouble seeing the
problem that occurs when you shift the corss-correlations in advance
rather than after the solution (which was done).

 > 
 > > Thanks - I guess - for letting out more rattlesnakes.
 > 
 > Sorry- this one's probably low priority, since its probably
 > an infrequent usage of BPASS.

    I hope it was infrequent - otherwise users are getting errors and
not telling me.  But the autocorrelations are supposed to be usable in
vlbi for bandpass.

 > 
 > Also, another question about the effect of the center-of-array bug -
 > am I correct in thinking that you will get the right amplitudes, but
 > the wrong phases out of running CALIB using a sub-image?  I tried it
 > and it seemed the resulting SN table amps. were in fact identical, 
 > whereas the phases differed markedly.

If one uses calib in this case, the source will be moved to a new
location - that is "all".  Thus the amplitude solutions should be fine
but the phases will include a center shift.  If you try instead a
UVSUB everything will be wrong.

I am looking at a messy solution (more header parameters) to this
issue which will fix future images but not past ones.

Cheers,

Eric Greisen