search

ratt-ru / CubiCal

A fast radio interferometric calibration suite.
GNU General Public License v2.0
18 stars 13 forks source link

apply gains to autocorrs #436

Closed o-smirnov closed 1 year ago

o-smirnov commented 3 years ago

@bennahugo has requested that we apply gains to autocorrelations when forming corrected data.

At the moment we zero them "as a precaution" here, but I see no reason not to flag them with FL.SKIPSOL instead (as we do with baseline length selections right now) to have them ignored yet corrected.

JSKenyon commented 3 years ago

I have taken a brief look. While I do think that it will probably be ok to do as you suggest, it will also require changes to the compute_corrected kernels. At present, they loop over baseline with a strict p < q approach. We will need them to include p == q. If you think this is worthwhile, I can try to implement something tomorrow.

JSKenyon commented 3 years ago

This slipped my mind - should I give it a shot @o-smirnov?

o-smirnov commented 3 years ago

Yes, please! I am buried in admin at the moment.

You need to fix the kernels, take out the "precaution", and set FL.SKIPSOL on them instead.

JSKenyon commented 3 years ago

@o-smirnov @bennahugo See #437. I have tried it out locally and nothing seems to burst into flame.

o-smirnov commented 3 years ago

and nothing seems to burst into flame.

That's my definition of a successful pipeline run!

o-smirnov commented 3 years ago

Also when I don't get arrested.