IMFIT vs UVFIT

Got an image problem? Let us help!

Moderator: Mark.Wieringa

bnorfolk
Posts: 45
Joined: Fri May 17, 2019 3:56 pm

IMFIT vs UVFIT

Post by bnorfolk »

Hi again,

I'm trying to determine the integrated flux of my source.

Using:
imfit in=imagefile region=arcsec,box(-8,-8,8,8) object=gaussian

Produces an integrated flux:
3.5969E-05 +/- 1.8854E-05
Where, using:
uvfit vis=visfile object=gaussian
Produces an integrated flux:
6.0837E-05 +/- 3.23E-05

Now I know that imfit will be looking at a specific region and should be lower, but if I unset the region I get the reading 1.298 +/- 7.386.

I'm also aware that I can use the spar variable, but when I do this the resulting integrated flux is always the predicted flux I enter in the spar variable which is obviously wrong, I just can't choose my flux.

I'm wondering what the correct way to integrate my sources flux is?

Thanks again,
Brodie
Mark.Wieringa
ATCA Expert
Posts: 297
Joined: Mon Feb 08, 2010 1:37 pm

Re: IMFIT vs UVFIT

Post by Mark.Wieringa »

Hi Brodie,

In both your fits the value is twice the error, so no significant detection. Source fitting doesn't work well below 5 times the noise. If your source is really that faint, your best bet is probably to just get a peak flux and maybe an integrated flux by using kvis to draw a little box around it and hitting "s". You could try a few box sizes to see if you get consistent results and to get a feel for the errors.
Making a lower resolution image will increase the peak flux for extended sources.

With imfit the box size needs to be big enough to capture all of the source, plus a band around it to get the noise level. Again, try different box sizes to see how consistent you results are.
The uvfit value will be unreliable for faint sources too, especially if there are other sources in the field.

A screenshot of your image would help to diagnose issue.

Cheers,

Mark
bnorfolk
Posts: 45
Joined: Fri May 17, 2019 3:56 pm

Re: IMFIT vs UVFIT

Post by bnorfolk »

Hi Mark,

Thanks for the reply, I've attached said image.

In regards to making a lower resolution image, what does that mean exactly?

Cheers,
Brodie
Attachments
Screen Shot 2019-08-05 at 10.48.32 am.png
Screen Shot 2019-08-05 at 10.48.32 am.png (36.64 KiB) Viewed 7360 times
Mark.Wieringa
ATCA Expert
Posts: 297
Joined: Mon Feb 08, 2010 1:37 pm

Re: IMFIT vs UVFIT

Post by Mark.Wieringa »

Hi Brodie,

thanks for the image.
Judging from the image, your source is only 2-3 times the noise level - unless you have a-priori knowledge that it is in exactly that spot, you wouldn't be able to claim it as a detection (need about 5 sigma). I think, as I mentioned before, the best you can do here is give an approximate peak flux estimate (with large error bar). Fitting is not really an option until the source is >5sigma.

To make a lower resolution image you can use the robust parameter - the beam size should increase as you increase this from -1 to +1. Alternatively you can use fwhm to specify the size of the gaussian you'd like to convolve with. From the image there is only a little evidence the source is extended (could be due to noise), so playing with robust to try and increase the peak value is your best bet.

Cheers,

Mark
bnorfolk
Posts: 45
Joined: Fri May 17, 2019 3:56 pm

Re: IMFIT vs UVFIT

Post by bnorfolk »

Hi Mark,

Thanks for the reply, I thought the best way to measure a flux was to use natural weighting?

If I use the robust parameter any where between -2 and 2 miriad prints I'm using a uniform weighting, and I either get an image with stripe bands for closer to negative robust values or a lower signal image for positive robust values.

My invert task for the previous image I sent is:
invert vis=visfile map=mapfile beam=beamfile imsize=3,3,beam cell=5,5,res sup=0 options=mfs,double

Cheers,
Brodie
Mark.Wieringa
ATCA Expert
Posts: 297
Joined: Mon Feb 08, 2010 1:37 pm

Re: IMFIT vs UVFIT

Post by Mark.Wieringa »

Hi Brodie,

you are right that you get the most sensitive image (lowest theoretical noise), using sup=0. It will not have the smoothest beam pattern though, so depending on the other sources in the field and the calibration quality, it may be easier to see your source with robust weighting. Robust=2 is almost identical to natural weighting. Robust=-1 will give longer baselines more weight, so if the calibration on those baselines is worse, you may see larger errors (worse dynamic range). This may be the case for high frequency observations.

If you are already using sup=0, the only way to increase the beam size further is to use fwhm - set it to something similar to your beam size or larger.

Cheers,

Mark
bnorfolk
Posts: 45
Joined: Fri May 17, 2019 3:56 pm

Re: IMFIT vs UVFIT

Post by bnorfolk »

Hi Mark,

Thanks again,

From my understanding the beam size changes in relation to the weighting, is there a way to determine its size previous to invert for its input into the invert task?

Cheers,

Brodie
Mark.Wieringa
ATCA Expert
Posts: 297
Joined: Mon Feb 08, 2010 1:37 pm

Re: IMFIT vs UVFIT

Post by Mark.Wieringa »

Hi Brodie,

not exactly, since it depends on the details of the observation (uv coverage). But after you've made you first image and cleaned it, you know the beam size for those settings and you can estimate from there. A very rough estimate is resolution ~ lamba/Dmax, i.e., observing wavelength divided by maximum baseline length. Uniform/robust<0 may be less than this, robust >0 or natural weighting will often be larger than this.

Cheers,

Mark
bnorfolk
Posts: 45
Joined: Fri May 17, 2019 3:56 pm

Re: IMFIT vs UVFIT

Post by bnorfolk »

Thanks heaps!
bnorfolk
Posts: 45
Joined: Fri May 17, 2019 3:56 pm

Re: IMFIT vs UVFIT

Post by bnorfolk »

Hi Mark,

I've implemented the fwhm parameter into invert for some of my faint sources. I received a higher peak flux but also a higher rms, and subsequently a lower S/N. Does this suggest my former method produced a more accurate reading? Or can I trust this higher flux reading?

Cheers,
Brodie
Post Reply