problems with running gpcal on CABB data at 16 cm

[dominics]

Hi,

I've heard that we still need to split up the wide-bandwidth CABB data at 16cm for polarization calibration, so that gpcal can run its course. Is this correct, and if so, could you explain in a few lines why this is?

cheers,
Dominic

[ste616]

Hi Dominic,

Yes that is correct. The problem here is that gpcal has no concept of a frequency-dependent gain. For normal gains, a combination of time-dependent frequency-dependent bandpass solutions and time-dependent frequency-independent gain solutions will work fine, even over a large fractional bandwidth.

However, for leakage calibration, this doesn't work. This is mostly a receiver issue, and the leakages will always be heavily frequency dependent. For small fractional bandwidths, the leakages will not vary much, but they will for a large fractional bandwidth. More simply, you shouldn't assume that the leakages that are computed for the "centre" of the band are valid for the entire band.

We currently feel that it is safest if you split your wide bandwidth data into 128 MHz chunks before calibration with gpcal and beyond. (If you're not interested in polarisation calibration, you can still calibrate reasonably accurately across the entire 2 GHz band in a single hit.) You should check this for yourself - you should see the calculated leakage parameters vary quite a lot between the chunks.

Sorry, that was a bit more than a few lines, but I hope it is clear.