So what did the wobbly G-terms do to our image?

<<Previous Up Next>>

So what did the wobbly G-terms do to our image?

Logged on 10/02/13 20:07:08

The easiest way to see this is to look at the statistics of the two cleaned images we made. The stats dervided by Purr when it attaches FITS files to a log such as this one will suffice, and these are available in the data products of the previous two sections.

The 'data range', the 'clip range' and the image thumbnails can provide a pretty decent picture of what's happened.

The data range values simply list the minimum and maximum values of the image in map units (Jy/beam, typically). For the first uncorrupted case the data range has an upper value equal to the intrinsic flux of the source. The feature in this map accurately reproduces its true flux. Compare this to the upper data range value in the second image. This peaks out at 0.095 Jy/beam, i.e. the source has been attenuated to below 10% of its true value!

Consider the screengrab of the Real part of the complex gain corruptions above. Although the sky is constant, the differing gain drifts in each station mean that the each element in the array sees an artificially attenuated (or amplified) signal that drifts in time. Remember the noise section in 'Simulations 101': making a map averages the many visibility measurements, and in this case we recover some 'average' flux that is way off the true value.

Remember also that we have attempted to deconvolve this source, albeit not very carefully, just 1000 blind iterations of clean over the whole field. Still, this should have been enough to remove the PSF sidelobes from the regions away from the source. This is where the clip range values and the image thumbnails demonstrate the other horrid effect of varying G-terms. The clip range shows the upper and lower pixel values that have been applied to render the thumbnail image.

For the first simulation with no gain corruptions the clip range spans 0.8 mJy to 3 mJy, and no background structure is visible. If you examine your FITS image here in Tigger you can see that the background structure is extremely low-level; it essentially results from the deconvolution imperfections. For the second, corrupted simulation the range is -8.9 mJy to 10 mJy. This range obviously covers much brighter emission and as can be seen from the thumbnail there is plenty of it. This emission is a fundamental 'noise' limit in this simulation. Deconvolution can't take care of it, only calibration can.