Calibration of data with correlator block dropouts

[and460]

Hi guys,

I'm calibrating some 16cm CABB data which was taken over a period where the correlator blocks dropped out in both IFs mid-observation. The observer appears to have stopped the obs, re-booted the blocks in question and gone back to the primary cal to set up again (dcal, pcal, acal etc), but has not taken any data on the primary after setting up. The observations were then re-started and proceeded as normal.

Does this mean the observations from this point on are pretty much useless though, because there is effectively no good primary cal data after the correlator re-start to GPBOOT etc. against to get the flux scaling correct in the calibration?

Cheers,

Craig.

[len067]

Hey Craig,

I'm just thinking about what this would do to the data. I guess you could have a re-scaling as a result of the acal (although I would assume this would be quite small) or you may get a phase slope across the band if the delays (for whatever reason) changed between the two dcal set-ups. The former should get fixed by using the calibrator source that you regularly visit. The latter might be a bit more tricky.

Have you tried to calibrate using the initial calibrator observations and then have a look at the regularly observed calibrator to see if the levels and phase slopes across the band look OK after the second set-up calibration? If they look similar to the observations before the problem occurred then I would expect that it should all just work fine.

Cheers,

Emil.

[and460]

Hi Emil - thanks for the reply. In regards to the phase: You mean use uvplt with something like axis=freq,phase to check the phase slope of the cal both before and after the restart?

[Mark.Wieringa]

Hi Craig,

phase slopes can be taken out by gpcal with nfbin>1, so I think simply running gpcal on the secondary with nfbin>1 should fix both amplitude jumps and phase slopes across the restart. You could specify a gpboot time range in the first section to avoid taking the data after the jump into account for the average secondary flux.

uvspec can show phase slopes.

Cheers,

Mark

[len067]

Hi Craig,

Yes, I was thinking of something like uvspec or uvplt over a selected period of time. (Thanks Mark)

Cheers,

Emil.

[and460]

Great - thanks for your time and help guys! I've run the cal and the data look OK to me...

Cheers,

Craig.

[ste616]

Hi Craig,

What will have happened is that when the observer did another acal, the correlator may have arrived at a different value for the strength of the noise diodes than it had before the glitch. The gains will be scaled by the ratio of the noise diode strength before to after the glitch, on the antennas that had their value changed. This is why doing an acal actually works.

So long as the correlator was looking at the same source both times an acal was done, the change should be small. You could probably see this jump if you used varplot to look at the xcaljy and ycaljy header parameters (Mark will correct me if I've gotten those names wrong). If the jumps are small enough, you may choose to forgo any complicated calibration shenanigans and accept a slight error in the flux density.