Page 1 of 1

How to produce the noise map?

PostPosted: Sat Dec 10, 2016 5:31 am
by taianmoon
Hi,

I am trying to produce a noise map with one of our 5.5GHz continuum observations. There is a sfind.rms image produced after the miriad task sfind to detect sources, but that map just has a constant value over the whole image (for all rmsbox values I tried: 50, 1000, 5000, 10000, 23494); also, the sfind.log catalogue has a rms(bg) column with all values being the same, which are weird...

I am wondering if there's a problem with that noise estimate and is that the right way to produce a noise map? I'm thinking of the residual map, but that might be different with the noise map?

Thanks!

Tai-An

Re: How to produce the noise map?

PostPosted: Mon Dec 12, 2016 3:16 pm
by Mark.Wieringa
Hi Tai- An,

I was hoping someone else would answer this one, as I've never run sfind before.
However when I ran it, with only 2 parameters set: sfind in=myimage options=rmsimg
I get an output list of 19 sources and an rms image that is not constant - there are peaks at the location of the strongest sources and small variations from block to block.

Make sure you are giving it the restored image (not a clean model).

If you're still having problems, post the task inputs so we can try to reproduce the problem.

Cheers,

Mark

Re: How to produce the noise map?

PostPosted: Fri Dec 16, 2016 12:45 am
by taianmoon
Hi Mark,

Thanks for the suggestion, It turns out that it's the "rmsbox" which I set it too large... But if I set rmsbox the default value (20 pixels), it works for one of our sources (G345) but not for the other (G018). For G018, miriad returns an error message "### Fatal Error: No convergence in GCF" at the beginning of the FDR analysis. Fortunately it work for rmsbox=40 and the resultant image is shown in the attachment (left is the restored image; right is the rms map). I wonder if there is any rule to decide the rmsbox value to input? There's this sentence from miriad manual but I found it confused to interpret practically: "It is suggested that this (rmsbox) be several to many times the beam size to prevent sources from artificially skewing the background estimates."

Thanks a lot!

Tai-An

P.S. The sfind task inputs that I use:

Task: sfind
in = (restored map)
type =
region =
xybin =
chan =
slev =
levs =
range = 0,0,log
cutoff =
rmsbox = 40
alpha =
xrms =
device = /xs
nxy =
labtyp = hms,dms
logfile =
options = normimg,rmsimg,fdrimg,mark,pbcorr,psfsize
csize =

Re: How to produce the noise map?

PostPosted: Mon Dec 19, 2016 8:01 am
by Mark.Wieringa
Hi Tai An,

great you got it working.
I think the GCF error is due to a failure to fit the background noise distribution - this may be due to a large source (compared to the rmsbox size).
I agree the instructions for setting the rmsbox size are a bit vague - I think that it is trying to find a balance between fitting variations in the noise across the field (rmsbox not too big) and not being affected by individual sources (rmsbox not too small).

Cheers,

Mark