General calibration recipe + nfbin questions

Got a calibration problem? Discuss it here.

Moderator: Mark.Wieringa

Post Reply
tg1989
Posts: 17
Joined: Sun Dec 07, 2014 3:04 pm

General calibration recipe + nfbin questions

Post by tg1989 »

Hi all,

Another few questions. Hopefully they are straight forward and not to silly.

I have data of twenty sources with some 11 phase calibrators, where one of the phase calibrators is 1934. A typical calibration solution I have been using is:

Code: Select all

mfcal vis=1934 refant=3 interval=0.1
gpcal vis=1934 refant=3 interval=0.1 nfbin=4 options=xyvary
gpcopy vis=1934 out=<secondary>
gpcal vis=<secondary> interval=0.1 nfbin=4 options=xyvary,qusolve refant=3
gpboot vis=<secondary> cal=1934
My first question, in the case where 1934 is not only used for bandpass/flux density calibration but also phase, what is the best way to calibrate the data? Is there any need to break the data into two parts - one for the bandpass/flux density where there was 10 minutes of straight time on source, and the second for each of the phase reference scans? Or is it appropriate to simply use step 1 and 2 from above, except with options=xyvary,qusolve in gpcal? Could I make this the 'default' calibration, or would it do something bad to the other 10 secondary phase calibrators? My first thought is that breaking the data into two files would make the most sense, but if gpcal corrects for phase changes through time it wouldn’t matter whether they are broken up or not.

Additionally, I was hoping to get clarification on how nfbin determines the frequency bins when a parameter for it has been supplied. I ask because, as mentioned in the previous thread, I have had to flag the upper end of my 11GHz data. When I supply an nfbin parameter, is it smart enough to compute these bins based on the minimum and maximum frequency of unflagged channels? I would think so simply because gpcal didn’t fall over, but i Would like to be sure. Perhaps including the frequency range of each bin could be included in the output where appropriate as it may be a useful reference?

And finally, I am curious if there are any tasks or ways to inspect the flags file of a uv-file so as to identify the channels which have been flagged - ideally only those which have been flagged for all records? I ask because I plan to divide my data up into multiple sub-bands using the line parameter in invert, and I hope to balance them such that they all have a consistent number of useable channels. Knowing how to do this will definitely help me remain consistent, especially if I try to automated the imaging aspect of the project.

Thanks (again)
Tim
ste616
Site Admin
Posts: 220
Joined: Thu Feb 04, 2010 3:27 pm
Location: Paul Wild Observatory Narrabri NSW

Re: General calibration recipe + nfbin questions

Post by ste616 »

Hi Tim,
My first question, in the case where 1934 is not only used for bandpass/flux density calibration but also phase, what is the best way to calibrate the data? Is there any need to break the data into two parts - one for the bandpass/flux density where there was 10 minutes of straight time on source, and the second for each of the phase reference scans? Or is it appropriate to simply use step 1 and 2 from above, except with options=xyvary,qusolve in gpcal? Could I make this the 'default' calibration, or would it do something bad to the other 10 secondary phase calibrators? My first thought is that breaking the data into two files would make the most sense, but if gpcal corrects for phase changes through time it wouldn’t matter whether they are broken up or not.
When you do the "gpcal" on 1934-638, you are calibrating it as a phase calibrator. The only reason why you don't have to give the "options=qusolve" parameter for 1934-638 is that we think we already know its linear polarisation to be zero, and so you don't need to solve for it (which is good since most people won't get enough parallactic angle coverage of 1934-638 to allow qusolve to work, although you might). After that step, you can use 1934-638 as a phase calibrator without doing anything further.

The only thing to possibly watch out for is selecting only a small range of 1934-638 data that overlaps closely in time and elevation to the source you want to flux calibrate. This is done in gpboot if you want to select the time range of the source you're calibrating, and it uses the entire time range of 1934-638. If this isn't appropriate, you might want to take a subset of the 1934-638 data using something like uvaver as the flux calibrator, but remember you can't do that after you've calibrated the whole 1934-638 dataset, since gpboot requires you to have the calibration tables in the dataset, not applied to the data. Perhaps it might be easiest simply to execute your idea of breaking the data into two parts.
Additionally, I was hoping to get clarification on how nfbin determines the frequency bins when a parameter for it has been supplied. I ask because, as mentioned in the previous thread, I have had to flag the upper end of my 11GHz data. When I supply an nfbin parameter, is it smart enough to compute these bins based on the minimum and maximum frequency of unflagged channels? I would think so simply because gpcal didn’t fall over, but i Would like to be sure. Perhaps including the frequency range of each bin could be included in the output where appropriate as it may be a useful reference?
I'm afraid nfbin just divides up the band without any regard to flagging. In most circumstances though, you probably don't need to even do "nfbin=4"; almost all of my calibration works well with "nfbin=2".
And finally, I am curious if there are any tasks or ways to inspect the flags file of a uv-file so as to identify the channels which have been flagged - ideally only those which have been flagged for all records? I ask because I plan to divide my data up into multiple sub-bands using the line parameter in invert, and I hope to balance them such that they all have a consistent number of useable channels. Knowing how to do this will definitely help me remain consistent, especially if I try to automated the imaging aspect of the project.
Use the "uvfstats" task with "mode=channel".
cheers
Jamie Stevens
ATCA Senior System Scientist
Mark.Wieringa
ATCA Expert
Posts: 297
Joined: Mon Feb 08, 2010 1:37 pm

Re: General calibration recipe + nfbin questions

Post by Mark.Wieringa »

Hi Tim,

On question 2:

nfbin divides the <selected> channels into that many bins, but it does not take flagging into account. If you want to restrict it to a range of channels, specify those with the line parameter. The mean frequency for each bin is stored in the solution tables, you can see them with gplist.

Cheers,

Mark
tg1989
Posts: 17
Joined: Sun Dec 07, 2014 3:04 pm

Re: General calibration recipe + nfbin questions

Post by tg1989 »

Awesome, thanks for the help! I had no idea about uvfstats either! Your explanation of the selected data while using nfbin does make some sense. For my data where I have flagged the upper the upper 800 channels, I see that in gplist there are only 3 frequency bins listed even though I specified 4. I guess this is because the last bin had no available channels available so no solution was derived. Really, I should have used the line parameter in gpcal if I really want the 4 bins, but as Jamie said this many is not always needed.

Thanks again Jamie/Mark!
Post Reply