mfboot and CABB data reduction at 6cm

[minh25]

Hi,

I reduced 6cm CABB data over a year ago following the recipe at:
http://www.atnf.csiro.au/computing/soft ... ode95.html

I am now looking at polarization, revisiting calibration to see if I can get any improvements with the nfbin option.

My steps are:

mfcal vis=1934-638.5500 refant=4 interval=0.1
gpcal vis=1934-638.5500 refant=4 interval=0.1 options=xyvary nfbin=4
gpcopy vis=1934-638.5500 out=0346-279.5500
gpcal vis=0346-279.5500 refant=4 interval=0.1 options=xyvary,qusolve nfbin=4
gpboot vis=0346-279.5500 cal=1934-638.5500
mfboot vis=1934-638.5500,0346-279.5500 select=source\(1934-638\)
(I use delhd to get rid of previous cal solutions before I run these steps)

But I see that the mfboot step scales the data by 1.000 (not too much of a worry) but adjusts the slope by 0.000 (possibly a worry). The largest adjustment I see over my 14 days of data is 0.002. Do we expect the spectral slope adjustment to be this small?

Do I need to gpcopy the solutions from 0346-279 to 1934-638 just before the mfboot? If so this should be put into the online manual.

cheers,
Minh

[ste616]

Hi Minh,

Since your 1934-638 data will already have the proper spectral slope, and mfcal will have used that information while constructing the bandpass solution, it is not surprising that mfboot will not have to correct the spectral slope by any significant amount.

Since gpcal with nfbin could potentially alter the spectral slope of the calibrator it is run on (with the exception of 1934-638, for which gpcal knows the flux), you might expect that mfboot could make some correction. But mfboot does not even look at data from any set that does not match the "select" criteria you give it. To look at the effect that nfbin has on your phase calibrator with respect to spectral slope, you would indeed need to copy its calibration solution to 1934-638 before running mfboot.

However, any correction made to the spectral slope by mfboot is done by altering the bandpass table. This would suggest that the bandpass is time-dependent, because mfcal should have made the "perfect" bandpass as it knows the actual spectral shape of 1934-638 and can shape the bandpass to ensure the shape is obtained.

That does not preclude that the bandpass is in fact time-dependent though. The mfboot spectral slope correction is quite a blunt instrument however, and there is a better way to account for this correction. In this case, I would run mfcal on your phase calibrator, and allow it to generate a time-dependent bandpass by using the "interval" keyword. You can then copy this calibration to 1934-638 and use mfboot, after doing your gpcal.

For example:
mfcal vis=0346-279.5500 refant=4 interval=0.1,1,1
gpcal vis=0346-279.5500 refant=4 interval=0.1 options=xyvary,qusolve nfbin=4
gpcopy vis=0346-279.5500 out=1934-638.5500
mfboot vis=1934-638.5500,0356-279.5500 select=source\(1934-638\)

Does that make any difference?

[Mark.Wieringa]

Hi Minh,

I think the simple answer is that gpboot now works in bins, so it can already correct any spectral slope differences between primary and secondary cal. After that there is nothing left to do for mfboot, so it gets numbers very close to 1 and 0. The mfboot step could be safely omitted from the recipe, but may still be helpful as a diagnostic for problems.

If there are time variations in the spectral response and you want to try and solve for these, follow Jamie's suggestion.
Another thing to watch out for is steps in the calibrated spectrum corresponding to the bins - if there is a large change in the opacity correction or in the RFI between primary and secondary, the gain solutions may be in error and neither gpboot nor mfboot will fix this.

Cheers,

Mark