Different spectral index on bandpass calibrator in 3mm band

[chu37330]

Hello all,

I got bandpass calibration problem when I was reducing 3mm data.
I followed the steps in this documentation, section "bandpass and flux calibration" and got different spectral index in 93GHz band and 95Ghz band.
https://www.narrabri.atnf.csiro.au/calibrators/calibrator_database_documentation.html

The following is detail of my calibration steps.
bandpass calibrator: 1253-055, flux calibrator: Mars

mfcal vis=bandpass interval=0.1 refant=2
gpcal vis=bandpass interval=0.1 refant=2 nfbin=2 options=xyvary,qusolve
gpcopy vis=bandpass out=flux
mfboot vis=bandpass,flux select=source(mars) device=/XS

And then I got negative and positive spectral index in 93GHz and 95GHz respectively.

How can I fix this problem?

Thank you all.

Regards,
Jimmy

[ste616]

Hi Jimmy,

Thanks for your question. It would be helpful for me to see some plots of your spectral shapes, just to see how much your spectral indices vary.

One thing you might want to try is using the select keyword, or clip keyword, in mfboot to remove any long baselines where Mars gets completely resolved out, as this will affect the flux density calibration and thus may cause this type of error.

From the mfboot documentation:

Key: select
Normal uv-selection parameter. This selects the data in the input
datasets to analyse. The data selected should consist of
a single planet or point source. See the help on ``select'' for more
information. For planets, you may wish to select just the shortest
spacing, where the planet is strongest.

Key: clip
For planets, this parameter can be used to discard data for baselines
that are significantly resolved. Data for a particular baseline will
be discarded when the expected flux density on that baseline is
less than the clip factor times the total flux density. The clip
parameter takes on values between 0 to 1.0, with the default being 0
(ie the default is to accept all data).

You should also include "mode=scalar" for your call to mfboot, to ensure that you aren't using one of the "triple" or "vector" modes, which won't be very useful for this.

[chu37330]

Hi Jamie,

Thank you for your reply.
Attached is the spectrum of bandpass calibrator after "mfboot"



I will try your suggestion when doing mfboot.

Thank you
Jimmy

[chu37330]

Hi Jamie,

I tried to do more selection when doing mfboot.
Something like:
mfboot vis=bandpass,flux select=source(mars),-ant(5) mode=scalar device=/XS
mfboot vis=bandpass,flux select=source(mars),-ant(1)(5),-ant(2)(5) mode=scalar device=/XS
or
mfboot vis=bandpass,flux select=source(mars) clip=0.2 mode=scalar device=/XS
mfboot vis=bandpass,flux select=source(mars) clip=0.5 mode=scalar device=/XS

I tried a lots combinations of anttennas and clip value from 0.0 to 1.0.
I also combined select=-ant() with clip in mfboot.

With these changes, spectral index didn't change a lot(from -1.185 to -1.262 for example).
I still can not get good result.

Do you have any idea?

Attached is flux density - baseline length of Mars in 93GHz band generated by mfboot.
Hope that this information will help.


Thanks.

Regards,
JImmy

[ste616]

Hi Jimmy,

Have a read of the bandpass and flux calibration section of the calibrator database manual.

http://www.narrabri.atnf.csiro.au/calib ... alibration

I think in your case, it is a little surprising that the spectral indices are different in each band, given how bright 1253-055 and Mars are, but there is always a chance that the slopes can be maladjusted by Miriad when using nfbin=2.

I would thus suggest trying the calibration again without the "nfbin=2", which might produce a slightly nicer result to begin with, and then continue by measuring the effective spectral behaviour of 1253-055 with uvfmeas, and using that to redo the bandpass calibration.

Let me know how you go!

[chu37330]

Hi Jamie,

After redoing calibration without nfbin=2, the results are similar, but the spectral shape becomes flatter in 93GHz band.

The following are calibration steps in detail and spectrum of 1253-055 in 93GHz and 95GHz
mfcal vis=bandpass interval=0.1 refant=2
gpcal vis=bandpass interval=0.1 refant=2 options=xyvary
gpcopy vis=bandpass out=flux
mfboot vis=bandpass,flux select=source(mars) mode=scalar device=/XS
I also tried select=-ant(X) in mfboot, but result did not change a lot.



Regards,
Jimmy

[ste616]

Hi Jimmy,

Getting better!

But to help any further I think I'm going to have to see your entire process from atlod onwards. Perhaps it would be easier also to let me know which raw data files you're using: I can do the reduction the way I normally do and see what happens.

[chu37330]

Hi Jamie,

The following are the commands I used from atlod.

atlod in=XXX out=YYY options=birdie,xycorr,rfiflag,noauto
atfix vis=YYY out=ZZZ
uvsplit vis=ZZZ select=-shadow(22.5)

uvflag vis=bandpass select=ant(6) edge flagval=flag options=brief
uvflag vis=flux select=ant(6) edge flagval=flag options=brief
uvflag ...

mfcal vis=bandpass interval=0.1 refant=2
gpcal vis=bandpass interval=0.1 refant=2 options=xyvary
gpcopy vis=bandpass out=flux
mfboot vis=bandpass,flux select=source(mars) mode=scalar device=/XS

Data I used is from project C3169 which is still restricted.
How can I give you the raw data of calibrator?

Thanks for help.

Regards,
Jimmy

[ste616]

Hi Jimmy,

Thanks for that, that all looks pretty good. I have access to all data, whether restricted or not, so I'll go have a look.

[ste616]

Hi Jimmy,

I think I've had a bit of success in calibrating this data, but it does require using the technique I describe in the calibrator database documentation. The concept is that you first calibrate without consideration of the spectral index, so making the assumption that the average flux densities of the two bands will calibrate to be very close to the correct value, then using that to determine the spectral index, then calibrating again with that assumed model.

In this case, we can calibrate 1253-055 as normal (you should start with an uncalibrated dataset, and I use 93000 below but everything should be done for both frequencies):

Code: Select all

mfcal vis=1253-055.93000 refant=2 interval=0.1

Then copy this to Mars and bootstrap the flux density scale:

Code: Select all

gpcopy vis=1253-055.93000 out=mars.93000
mfboot vis=mars.93000,1253-055.93000 "select=source(mars),-ant(1)" mode=scalar options=nospec clip=0.5


I excluded antenna 1 from the boot because the baselines that included it were responsible for those higher-than-expected fluxes in your mfboot plots. This is what my plot looks like from 93 GHz:

mfboot.93000.png (6.75 KiB) Viewed 1868 times

With that done, we can measure the flux model for 1253-055:

Code: Select all

uvfmeas "vis=1253-055.*" stokes=i order=1 options=log,plotvec,mfflux

I get a flux density model of 13.6337 Jy @ 93 GHz and a spectral index of 1.3597. Here's what the first round calibration looks like.

round1_1253-055.png (10.87 KiB) Viewed 1868 times

We can now redo the calibration. Again, from an uncalibrated stage:

Code: Select all

mfcal vis=1253-055.93000 refant=2 interval=0.1 flux=13.6337,93.0,1.3597
gpcal vis=1253-055.93000 refant=2 interval=0.1 flux=13.6337 spec=93.0,1.3597


This gives the result:

round2_1253-055.png (10.83 KiB) Viewed 1868 times

Which is much better, but not much good unless we can also get a good result using this calibration. So let's see what happens with the gain calibrator:

Code: Select all

gpcopy vis=1253-055.93000 out=j1534-5351.93000
gpcal vis=j1534-5351.93000 refant=2 options=xyvary,qusolve interval=0.1
gpboot vis=j1534-5351.93000 cal=1253-055.93000


This gives us:

round2_j1534-5351.png (12.3 KiB) Viewed 1868 times

Which looks pretty good to me. What do you think?