The ATCA Forum

Hi there,

I'm imaging CABB data at 2.1GHz so I've decided to image them in 500MHz chunk to make mfclean tolerate the bandwidth. Now I need to match the resolution and use linmos to generate final 2GHz bandwidth. I'm using Convol to match the resolution of my 3 cleaned images with the lowest resolution I have(ch,500,1500) but it looks like that these 3 convolved images dont have the same resolution as the lowest one.

Is there any way to match the resolution in CABB data for 500MHz images.

Cheers,
Sara

Hi Sara,

convol with the input beam and options=final specified should do a reasonable job of getting the resolutions equal. Of course doing this will effectively throw away about half the data at the high frequencies, but that cannot be avoided. What resolutions are you getting?

Another approach would be to specify fwhm in invert to match the beam at the lowest freq, with uniform weighting this should also get you close. A bit of trial and error may be needed and an exact match will not be possible, but if they are close enough you could use the average of the beam parameters as the final beam.

Cheers,

Mark

Hi Mark - just a quick question about using the fwhm parameter in INVERT which follows on from Sara's post and your reply.

From the documentation it seems a little unclear what specifying a value for fwhm actually does. The help file says:

"It specifies the FWHM of an image-domain gaussian -- tapering the visibility data is equivalent to convolving with this image-domain gaussian."

From trial and error, it seems that setting the fwhm param effectively specifies the approximate final resolution of the dirty image that INVERT produces. Is this correct? The sentence from the help file that I pulled out above though seems to suggest that a dirty image is first produced by INVERT at the 'native' resolution, and is then convolved with a 2D Gaussian of width specified by the fwhm parameter, which would lead to a frequency-dependent resolution once again. Perhaps I'm just reading it incorrectly, but it did trip me up for a while thinking about it so I thought I'd ask and get the definitive answer!

Cheers,

Craig.

Hi Craig,

Yes you are correct - it attempts to specify the output resolution. What it actually does is apply a radial gaussian taper to the visibilities, which as it claims is equivalent to convolving with a gaussian in the image, but a lot less work. The width of the taper in the uv plane is calculated from the fwhm in arcsec specified and the average frequency.

If your data has plenty of visibilities in the uv range where the taper drops off, the result should be an image with close to the expected resolution.

Frequency dependent resolution? Both convolution and tapering would suffer from this in the same way. The 'native' resolution of the data is convolved with the gaussian to produce the output resolution. For mfs imaging this should be fine, for cubes I think a frequency scaling fudge is done so all planes have the same beam.

Cheers,

Mark