Accuracy of CONVOL flux normalisation

Is MIRIAD being a pain? Let us know your experience.

Moderator: Mark.Wieringa

Post Reply
and460
Posts: 79
Joined: Thu Sep 01, 2011 12:23 pm

Accuracy of CONVOL flux normalisation

Post by and460 »

Hi guys,

I have a weird problem. I have a few hundred sources in a large mosaic observation. I have extracted Stokes I flux densities at 8MHz intervals over 1-2GHz, after having used CONVOL to convolve the mosaic images at each frequency to a common resolution (that of the lowest freq mosaic image). I fit a model to the Stokes I data for each source, and then calculated the spectral index. Separately, the data was imaged into a giant MFS image, and the spectral index was calculated for each source using MFSPIN after applying MFCLEAN.

Now, the distribution of spectral indexes calculated using each technique looks similar for the source population, but there is an offset of about 0.3 between them - i.e. the SI for sources calculated using the first method seem to be, on average, about 0.3 flatter than the SIs calculated using the second method. Having a bit of a think about what could be going wrong (given the same data set is being used in both instances), the only thing I could think of was that CONVOL was applying an erroneous, frequency-dependent flux normalisation during the convolution process. I calculated that I'd get the observed 0.3 flattening in SI if the flux normalisation errs by about 20% of Stokes I at 2GHz compared to that at 1GHz.

So I guess my question is: how likely is it that I'm on the right track here? Is it likely that the CONVOL flux density correction could err by this amount, or is this highly unlikely?
Mark.Wieringa
ATCA Expert
Posts: 297
Joined: Mon Feb 08, 2010 1:37 pm

Re: Accuracy of CONVOL flux normalisation

Post by Mark.Wieringa »

Hi Craig,

I think convol is probably doing the right thing, using gaussians for all the beams, it is just applying the beam area correction. It doesn't look at the frequency in the header.
There are plenty of other things that could be wrong though:
  • * the primary beam model is inaccurate near 2 GHz, the actual beam is wider than the standard miriad model (a new model exists but you need to fiddle with the image headers to use it)
    * the effective frequency in a wide band mosaic is lower as you move away from the field centre (you can make an image of this with linmos)
    * mfspin and mfclean can't really cope with a fractional bandwidth >40%
Tom Franzen noticed earlier from a plot of the median spectral index vs distance that mfspin tends to overcorrect for the primary beam as you move away from the pointing centre.

It might be instructive make a similar plot for your two cases and see if a pattern emerges.
Does the median spectral index of your sources give a clue as to which method gets closer to the truth?

Restor and linmos can now use the spectral index plane in an attempt to do a better job at the correction.

Cheers,

Mark
Post Reply