Polarisation imaging

[twshimwell]

Create Q and U cubes with 480 channels about 3 degrees across. This is ~7000 pixels with 1.5arcsec cellsize (current high imaging cellsize) or ~2400 pixels with 4.5arcsec cellsize.

Testing so far has shown that ~1000 pixels is fine with the memory constraints but its unclear what the maximum is.

It would be also nice to produce a stokes V image. My thoughts would be that this can be done with the 4.5arcsec cellsize at 20arcsec resolution.

[sposullivan]

The RM synthesis products will be larger (depending on sampling in Faraday depth). But at least we would want to cover -150 to +150 rad/m2, and a sampling of 0.15 rad/m2 would be good, which is ~92 GB cube. And we need 3 (p, q and u) cubes. And that is just the dirty cubes, not doing RMclean. We could bring up any long-term storage issues at the Krakow MKSP meeting?

[twshimwell]

OK that sounds a bit large then... Presently we have the primary beam corrected and uncorrected cubes stored. Can we perhaps just keep the primary beam corrected ones?

[rvweeren]

On 6 Aug 2018, at 12:24, twshimwell notifications@github.com wrote:

OK that sounds a bit large then... Presently we have the primary beam corrected and uncorrected cubes stored. Can we perhaps just keep the primary beam corrected ones?

For polarization it can help to have uncorrected (for the primary beam) cubes. That makes it easier to deal with the Ricean bias in the polarization analysis.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[mhardcastle]

Also if we ever want to make mosaics in Q and U we want the beam correction.

[cyriltasse]

Why not keep the RM cubes only on islands?

[sposullivan]

Maube sensible for field sources, not diffuse. Might effect anyone wanting to do polarization stacking later... For 1 mJy/beam cut of some pointing I was testing, goes from 5.8e6 pixels to 8e4.

[cyriltasse]

But the diffuse stuff is the low-res cube Marta was testing, or some kind or some kind of intermediate scale where you'd need the high res cube?

Le lundi 6 août 2018 à 17:18:23 UTC+2, sposullivan <notifications@github.com> a écrit :  

Maube sensible for field sources, not diffuse. Might effect anyone wanting to do polarization stacking later... For 1 mJy/beam cut of some pointing I was testing, goes from 5.8e6 pixels to 8e4.

— You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHub, or mute the thread.

[sposullivan]

Yes just low-res (4') and high-res (20") cubes at the moment, as far as I know. Little RM cubes around every bright Stokes I source at 20" would be really cool as a pipeline product, linked to a VO catalog table maybe.

[mhardcastle]

I think we've decided to live with the large datasets for now or to compress as suggested in #90. So closing this.