Page 1 of 2

Imcat and imcomb error

PostPosted: Wed Apr 24, 2013 3:17 pm
by saracmbr
Hi there,
I need to make cube image and I need to include whole 2GHz bandwidth. When I want to clean this cube, the clean and restor will use the beam has been made at the central frequency of 2.1GHz. I dont think I can use same beam for all 2048 channels so I need to do this in chunks and then concat the images. I tried to image 2 single channel to try imcat and imcomb but there is no success so far.
The error of imacat is : Images are not contiguous on axis 3.
The error of imcomb is : Geometry of image 1 differs from image 2, can not interpolate.

Why these errors rise and how I can make a cube with the proper beam to deconvolve.

Many Thanks,
Sara

Re: Imcat and imcomb error

PostPosted: Wed Apr 24, 2013 3:36 pm
by ste616
Hi Sara,

If you make a cube in invert (ie. do not use options=mfs), then invert will compute the beam separately for each of the planes, and clean and restor will use these beams appropriately.

Re: Imcat and imcomb error

PostPosted: Wed Apr 24, 2013 3:42 pm
by saracmbr
Is this true for mosaic as well?

Re: Imcat and imcomb error

PostPosted: Wed Apr 24, 2013 3:53 pm
by ste616
It's true for any cube made by invert.

Re: Imcat and imcomb error

PostPosted: Mon Apr 29, 2013 11:57 am
by ste616
Hi Sara,

After some discussions with Mark Wieringa, I must apologise and correct what I said about invert: it does indeed only make a single beam at the central frequency of the cube.

We'll look into the imcat and imcomb errors.

Let's talk about your issue though: why do you need the entire 2 GHz bandwidth for a spectral cube, especially at 2.1 GHz? Are you searching for some spectral lines that could be anywhere in that frequency range?

You don't necessarily have to make an image for each individual spectral channel: the 16cm band beam will not change that quickly. You could probably get away with cubes covering 16 MHz at the very least (the old ATCA correlator had that kind of bandwidth in spectral line mode), and I'd think you could double that to 32 MHz without too much issue. That at least takes the problem from making 2048 images to making only 128 cubes, and then you only need to make those cubes where you are expecting a signal.

Again, sorry about leading you astray here.

Re: Imcat and imcomb error

PostPosted: Tue Apr 30, 2013 1:00 pm
by Mark.Wieringa
Hi Sara,

you are getting the the imcat error message because the channel order is the reverse of the
frequency order. If you specify the images in reverse channel order imcat works fine.
It happens because the increment of the 3rd axis for a single channel
image comes out as +1MHz, it should really be -1MHz.

imcomb should not be used for this - it tries to merge the images into one.

Cheers,

Mark

Re: Imcat and imcomb error

PostPosted: Wed May 01, 2013 2:00 pm
by saracmbr
Hi,
I need all available channels to use in RM-synthesis code. I cant use averaging since it will spoil the calculations in that code. After removing those channels without any data left (100% flagged) I ended up to 1505 channels. I now have 1505 uv data needs to uvglue together but it seems that it take ages to do this.

In Imcat do I need to give the images from highest channel to lowest channel?

Cheers,
Sara

Re: Imcat and imcomb error

PostPosted: Wed May 01, 2013 2:19 pm
by ste616
Hi Sara,

I didn't mean that you have to average the data together. Rather, if you make a cube with 1 MHz channels but 16 MHz wide, each plane will be deconvolved with a beam that is at most 8 MHz away from its frequency. For most purposes, this should not change your results. And you can make a small bandwidth cube and compare it to the same cube made by individually deconvolving the planes to make sure the differences are acceptable.

Re: Imcat and imcomb error

PostPosted: Wed May 01, 2013 2:23 pm
by Mark.Wieringa
Hi Sara,

yes, for bands where the frequency decreases with channel number, like 16cm, specify the channels from high to low for imcat to work.

With uvglue you cannot have missing channels or the frequency labelling will be wrong in the output, the sequence of channels needs to be contiguous.

Cheers,

Mark

Re: Imcat and imcomb error

PostPosted: Wed May 01, 2013 2:34 pm
by saracmbr
Do I need to choose line when I'm imaging?
line=ch,1,1,16,1

Cheers,
Sara