search

GBTAmmoniaSurvey / GAS

observing scripts and files related to the GBT Ammonia Survey (GAS, PI: Jaime E Pineda & Rachel Friesen)
MIT License
6 stars 13 forks source link

Noisy footprints in some spectral channels after gridding #154

Open rfriesen opened 7 years ago

rfriesen commented 7 years ago

Need to include a mechanism to check for noisy scans in reduction pipeline.

low-sky commented 7 years ago

Maybe this will do?

rfriesen commented 7 years ago

Certainly looks reasonable! We can re-run one of the DR1 sets that shows the issue through the gridder and see if that solves it.

low-sky commented 7 years ago

Okay, I (or someone else) can PR this snippet in soon (I'm on the way out the door). Do you have a region in mind?

rfriesen commented 7 years ago

All the (3,3) cubes in the DR1 regions show this issue, so probably NGC 1333 is the smallest? Doesn't need to be done today, I just wanted to post this since you had mentioned you had a possible solution! I'm happy to look into it, later.

rfriesen commented 7 years ago

I'm testing the Tsys check as implemented in the KEYSTONE pipeline, but find the same issue in the resulting L1688 NH3 (3,3) cube. It's often (but not always) an entire footprint that has a strong positive or negative offset in a single velocity channel (see image).

@jpinedaf , didn't we see something similar in one of the Orion A footprints? How did you resolve that?

snapshot2

low-sky commented 7 years ago

Thanks, @rfriesen

I've definitely seen evidence for this as well. Do you happen to have a session / footprint number that has the glitch? I can track it down, but if you already know, it would make it faster.

The issue appears to be a high spike in part VEGAS. It should be flaggable in the time series. Since we pass to the FFT domain for spectrum shifting, there could be a flagging step there as well.

rfriesen commented 7 years ago

@low-sky I just took a look and the bad footprints for L1688 seem to all come from specific sessions: The large negative features are in L1688-3,-6, and -9, observed in Session 42. Positive/noisy features are in L1688-5 and -8, and were observed in Session 15.

low-sky commented 7 years ago

:+1:

Thank you. I'm working on GBT software stacks now.

low-sky commented 7 years ago

Addressed in #158

low-sky commented 7 years ago

Some progress: spikes are reduced but not eliminated in this approach. Baselines appear a lot better. This does a spectrum space flagging. I'll work on FFT flagging now that the grid-code refactor is approaching usability.

progress