[daip] Data combination

[Eric Greisen]

Danielle Fenech writes:
 > Thank you for the reply, I'll try and explain a little more of what I'm 
 > trying to do, I'm not sure I did a great job of it, but from some of what 
 > you said, I think the answer is going to be it's not possible.
 > 
 > To give you a little bit more information, I'm combining 4.5GHz and 
 > 6.5GHz observations and need to be able to extract, at a later point, 
 > spectral information from the singular combined dataset.
 > 
 > In order to do this I need the combined data to be able to distinguish the 
 > 4.5GHz data from the 6.5GHz data when either imaging or splitting the 
 > channels.
 > 
 > If I understand correctly reference frequency information is stored in the 
 > file along with information of the number of IFs, channels, channel widths 
 > etc. which imagr (for example) uses to calculate the appropriate u,v,w's.

         IMAGR recalculates u,v,w for spectral channels from the
initial u,v,w CORRECTLY computed for spectral channel 1.  The
correction is in second order wrong and that may matter with such a
wide frequency separation.  You really need to image the 2 bands
separately and then perhaps make a spectral index image and only then
perhaps image the 2 together with the spectral index image option.

Note too that the primary beams are vastly different at such wide
separations and so you are not even imaging the same sky.

 > 
 > During this process the reference frequency is taken to be that of the 
 > header and therefore of the first dataset in the combination (in this 
 > case 4.5GHz) so when the u,v,w's are calculated for the 6.5GHz portion 
 > of the combined data they are calculated using 4.5GHz as the reference 
 > frequency instead of 6.5GHz, so the u,v,w's are therefore incorrect.

        see above - if they were wrong by 50% you would see double
images everywhere.

 > 
 > The two sets of data have already been split from their original 
 > multi-source files with the calibration applied before combination but are 
 > still multi-channel data.
 > 
 > I know data combination is common but as far as I'm aware maintaining the 
 > ability to distinguish the frequency of the two portions of combined data 
 > in this way isn't usually the point.
 > 

One cumbersome way to solve the issue is to convert both data sets
into 2 IF data sets with the IF 1 at one frequency and IF 2 at the
other.  Each data set would be half flagged of course and so this
would double your data volume.  This may be tricky to do but there are
almost certainly tools to do so - if not a FUDGE clone could be
written.  Look at tasks UVCOP, BLOAT, VBMRG, VBGLU and there are verbs
to change table and header values too.

Eric Greisen