[daip] Re: combining frequencies

[Eric Greisen]

Tracy Clarke (Foreign National) writes:
 > Hi Eric,
 > 
 >  	I have sources in my field that probably run from flat spectrum to 
 > a spectral index of steeper than -2.5 so I am worried about the spectral 
 > effects. I am working with 321.5 and 328.5 MHz data at the moment so I 
 > shouldn't be getting too large of a flux variation (just a few precent). 
 > If I don't go with the DBCON approach does that mean I should go with a 
 > BLOAT and then combine them that way keeping all the spectral channels 
 > separate? What will happen with the intermediate region where there are no 
 > adjacent channels?

     They will be flagged out and IMAGR or whatever will handle them
fine so long as you are doing pseudo-continuum.  If you start doing
things that worry about frequency effects then the empty channels will
slow things down a lot.  IMAGR does handle different IFs correctly -
maybe a BLOAT to make IFs instead of spectral channels should be
written.

But BLOAT only works if the two data sets are identical in their
antennas, observing times, etc.  It was written for old correlators
that could only correlate a few channels at a time when many were
actually observed.  I suspect that you do not have this situation.
The two data sets do not have an identical pattern of times,
baselines, etc and DBCON is really the only approach.

IMAGR probably needs to be smarter to have subarray known too  - but
that would only make it all slower again.

 > 
 >  	Wendy mentioned the CORFQ followed by the DBCON but to me that 

       CORFQ is the worng thing to do.  You did not observe at a wrong
frequency so the u,v,w in your data set are correct for the observing
frequency.  If you change them then they will become wrong.

 > does not seem correct to me since it will not help with the 
 > multi-frequency imaging issues. Other suggestions were just straight DBCON 
 > (which I have been using in the past). The other side of this is that I 
 > have a student here pulling lots of archival data over and putting it all 
 > together so I decided it was really time to be sure we are doing things in 
 > the optimal way for taking advantage of the multi-frequency data sets on 
 > these sources that have large ranges of spectral indices. Cheers,
 > 

Eric Greisen