[daip] Re: question about DBCON

[Eric Greisen]

Claire Chandler writes:
 > Why does DBCON change the reference pixel from the centre of the band to
 > pixel 1 (and scale the u,v,w in the process)?  Wouldn't leaving it at the
 > centre of the band mean that there is less deviation in u,v,w across the
 > band from nominal?

This was a defense against the folks who SPLIT their data into single
channels (making the ref pixel be other than one for single freqs and
different for each) and then DBCONed tham all back together.  By
changing the u,v,w we get it right for the other samples and the whole
thing is okay.  Tim's code coould not handle multiple channels so this
was needed to take data to SDE.

The error in correcting u,v,w is non-existant if the data are at the
same frequency.  If they are at different frequencies, then there is
an error in how the second data set's u,v,w at the highest channels
are corrected.  It is true that we could reduce this by using the
center of real spectra.  This will matter for the EVLA but I wonder
how much DBCON of different frequencies we will see there.

Eric