[daip] subtracting 'calibrated' source models from UV data

[Eric Greisen]

Huib Intema writes:
 > Dear Eric,
 > 
 > I'm using AIPS tasks to subtract bright sources in large fields obtained 
 > from low frequency observations (<300 MHz). Each bright source generally 
 > has a unique calibration (SN table) that is valid in that particular 
 > viewing direction due to ionospheric phase disturbances. This peeling 
 > process involves many repetitions of the following call sequence:
 > 
 > 1. SPLIT to apply a given SN table to the UV data
 > 2. UVSUB/OOSUB to subtract a source model from the UV data
 > 3. CLINV to invert the SN table used in step 1
 > 4. SPLIT to apply the inverted SN table to the UV data resulting from step 2
 > 
 > I'm worried about applying and un-applying SN tables multiple times, as 
 > this introduces noise due to numerical round-off errors. A numerically 
 > more stable way is to generate a UV data set from the source model 
 > (through UVSUB/OOSUB opcode 'MODL'), apply the inverted SN table to this 
 > data set and subtract it (visibility by visibility) from the original UV 
 > data.
 > 
 > I'm aware of the AIPS task UVMTH, but this task does time-averaging on 
 > one of the UV data sets before combining two UV data sets, so this is 
 > not useable for my proposed scheme. As far as I know, what I propose is 
 > currently not possible in AIPS.
 > 
 > For a possible implementation of this functionality in AIPS I can think 
 > of two possibilities:
 > * Change in UVSUB/OOSUB so that one can apply an SN table to the UV data 
 > that is internally generated from the specified source model before it 
 > is subtracted from the specified UV data.
 > * Change in UVMTH so that one can subtract two UV data sets visibility 
 > by visibility without time-averaging, possibly allowing application of a 
 > SN table to the second UV data set (the one that is subtracted from the 
 > first).
 > 
 > Would you consider adding this functionality to AIPS. I would be most 
 > grateful.
 > 
 > Best regards

Would the task DIFUV suffice?

Eric Greisen