Continuum mosaic: Edge/noise problems between fields

[keflavich]

(fields ar (center) and k (bottom-right))

This looks to be caused by significant depth differences. "I’m just looking at the two weblogs, and field k has a lower RMS, significantly smaller beam, and > 2x aggregate bandwidth compared to field ar" from Dan

[d-l-walker]

Weblog summaries for continuum images

Field ar

Field k

[keflavich]

Right, that's a huge difference in sensitivity: 83 uJy/beam in 1.57x0.929 = 56 uJy/beam in a 2.02x1.55" beam, so about 2x better noise.

[keflavich]

This issue manifests in part as excess noise along the edges of peak intensity maps.

I had started working on debugging this at some point, but didn't track my work well:

https://github.com/ACES-CMZ/reduction_ACES/blob/4ed2804f883ddf148d02fbd18eb97812c078f5dc/aces/imaging/debug_mosaic_weights.py

[keflavich]

e.g., this:

[keflavich]

The problem was partly or entirely (not sure yet) that the weight images are a different shape from the feathered (12m+7m+TP) cubes, since the latter are cropped down to include only significant pixels.

I have possibly solved this with a combination of fixes to the ACES pipeline, spectral-cube, and reproject.

[keflavich]

This seems mostly solved, but not perfectly. Is it possible that the remaining problem is just that some fields are under-weighted? Maybe we should normalize the peak weights?

https://github.com/ACES-CMZ/reduction_ACES/blob/28a2bb4c8be3d51dcab7ca90c6c1c51edac416ad/aces/imaging/debug_mosaic_weights.py is the tool to test this.