Notes to the imaging, CLEAN and difference results for the C resolution
The results shown in this section have been obtained in the following fashion:
Due to problems in registering the original SIL images (which are 1024 pixels square with a 0.01arcsecond pixel scale) with the CLEANed images for the C resolution (which are 1024 pixels square with a 0.04 arcsecond pixel scale) it was decoded to try subsampling the SIL images to obtain versions at the same pixel scale. This was done using the SUBIM task with XINC and YINC = 4, and OPCODE = 'AVE' to average over the 4x4 pixels which become one in the resulting image. This gives a 256x256 image which is padded up to 1024x1024 using PADIM with IMSIZE=1024 and CPARM=0 to pad with pixels of value zero. If 'magic blanking' should have been used please tell me!
This having been done, the padded subsampled image was used as the model input to the imaging part of the simulation engine, with the resulting visibility/image/etc. data CLEANed as before to yield the image and CLEAN results shown here.
The differencing process requires the input model to be convolved by a gaussian equal to that used as the restoring beam in the CLEAN process. The BMAJ, BMIN and BPA values of this beam is specified as I have described earlier, and is applied to the input model using CONV, as for the A resolution results. However for the C resolution we have used a subsampled input model and must therefore recalculate the scaling factor necessary to get the convolved model images of comparable values to that of the CLEAN images. This uses the same formula as before (i.e pi d squared over 4 ln 2, where d is the diameter of the synthesised beam in number of pixels, obtainable using IMHEAD) and gives reasonable results for all the models except MPD. MPD was an anomaly also for the A resolution, with a scale factor of 2 used sucessfully in the end- this would imply that 2/16 would be appropriate here, but inspection of the values show that the scale factor to be used is 1! I assume this behaviour is some consequence of the MPD image being artificial, with no BMAJ, BMIN etc values in its FITS header. As for A the scaling factor seems slightly off, with the zero of the greyscale for the greyscale images not being centered exactly. I also apologise for different scales on the images- it is difficult to specify sensible ranges prior to seeing the results, so it seems better to first see what we get.
The scaled and convolved version of the (subsampled and padded) input model is now subtracted pixel by pixel from the CLEAN image to give the difference maps shown. Only the central regions of interest is shown, corresponding to BLC=384, TRC=640.