Removing the effects of sources in the field using Miriad
(by Demerese Salter --- 11/15/2011)
(edited by Shaye Storm --- 11/16/2011; updated Feb 2012)
A bright point source in the field (offset from phase center) can cause ripples across the image plane when Fourier transforming the visibility data. The closer to phase center, the lower the frequency of the ripples and the less likely the positive and negative parts of the ripples will cancel out, yielding noticable effects on the analysis of the science target. Extended sources do not impact the field as much because the different phases (from the different positional offsets) average out more quickly across the image. In general, the more beams away you are, the lesser the impact on your image (and the flux of your object at phase center, depending on whether it is located in a peak or trough of the ripples caused by the bright nearby source). The moral of the story is that if you have a bright source within a primary beam of your science source, you better do a good job of removing it from the field.
Removing Point Source
To remove a point source, use uvmodel. Here is an example removing a point source of 0.005 Jy at the given RA/DEC offset:
uvmodel vis=$data options=sub flux=0.005 offset=32.47,404.42 line=chan,240,1,1,1 out=$out
Removing More Complicated Sources
Sometimes you may have two bright sources near the phase center, for example, in the case of a (visual) binary. To individually analyze each member of the binary in the uv-plane, you need to remove the other source. These sources may not be point sources, so simply using uvmodel subtraction might not work! One method is to:
- 1. Invert and clean visibility data for full field with all the sources in it.
(*2. In clean space, use maths to set to zero all parts of the image EXCEPT the area containing the source you want to remove. OPTIONAL?)
- 3. Use imsub to make a subimage containing the sources you want to remove from the larger image -- let's call the subimage sub1.
- 4. Use uvmodel to subtract the sub1 image from the full visibility data by inputting it as the model parameter and using options=subtract to create a modified visibility dataset without that source.
- 5. Check that it worked by making an image (invert, clean, restor) on the modified visibility data.
- 6. Now you are ready to run uvfit or uvamp on the modified visibility data with the targeted source defined at the phase center using the appropriate offset or center keyword. Get of the offset from phase center by running "imfit" on the restor'ed image.
1. invert vis=crowded.cal map=crowded.map beam=crowded.bm imsize=129,129 cell=1,1 options=systemp,double robust=0 line=chan,1,1,39,39 clean map=crowded.map beam=crowded.bm out=crowded.cln gain=0.1 niters=1000 3. imsub in=crowded.cln region='box(27,30,54,52)' out=sub1.cln 4. uvmodel options=subtract model=sub1.cln vis=crowded.cal out=lesscrowded.cal 5. invert vis=lesscrowded.cal map=lesscrowded.map beam=lesscrowded.bm imsize=129,129 cell=1,1 options=systemp,double robust=0 line=chan,1,1,1,1 clean map=lesscrowded.map beam=lesscrowded.bm out=lesscrowded.cln gain=0.1 niters=1000 restor model=lesscrowded.cln beam=lesscrowded.bm map=lesscrowded.map out=lesscrowded.res fwhm=3.0 6. uvamp vis=lesscrowded.cal bin=12,10,klam offset=-0.435,0.335 > lesscrowded.uvamp.log (offsets derived from > imfit in=lesscrowded.res object=gaussian region='box(50,50,60,60)')
- Another method is to select the clean components from the u,v data using uvmodel and then subtract from the u,v data the bright, extraneous sources. ...????
- The task imhist' will give you the image statistics. ????
- If your image has a high dynamic range, then a Direct Fourier Transform (DFT) is more precise in removing point sources because you know exactly where the source is in the u,v plane. In contrast, a Fast Fourier Transform (FFT) takes advantage of gridding, reducing your positional accuracy in exchange for computational speed. The error in position means that you may not be fully subtracting your point source and its corresponding effects from the data. If you could, you would make a map, find all the point soruces, do accurate DFTs on them, subtract the point sources, remake the map, and repeat until you got high dynamic range.