search

ratt-ru / pyxis

Python Extensions for Interferometry Scripting
3 stars 5 forks source link

Zero values for the visibilities depending on simulation time #78

Closed alex-onose closed 8 years ago

alex-onose commented 8 years ago

When using lwimager.predict_vis I am getting strange behaviour for some of the uv points. The UV coordinates are correct but the associated predicted visibilities are set to zero.

In all simulations the first predicted visibilities are correct. Later on more and more are set to zero. I'm not very familiar with the tools so it might just be something obvious that i'm missing. There are no errors in the log files and everything seems to be performed successfully.

Here are some examples for the VLA for two simulations with different length in time. test_4

test_2_1

For the second one the error for the visibilities agains values estimated in Matlab is plotted below test_2_2_vis_error

In addition, running clean from lwimager seems to not take into account any of the points that are set to 0 by lwimager.predict_vis (even if I replace them in the MS file). Wsclean seems to use all the points when used on the same MS file

Any help with this would be appreciated.

o-smirnov commented 8 years ago

Very curious, looks like lwimager is simply ignoring some MS rows (since as you say they're also not taken into account in clean). Could be a data selection issue. Please paste the logs from the relevant lwimager and wsclean invocations here.

alex-onose commented 8 years ago

Here are the logs:

lwimager + casa log-pyxis.txt log-simms.txt

wsclean log-wsclean.txt

o-smirnov commented 8 years ago

Odd. From the logs it does not seem like a data selection problem.

Can you check if the missing visibilities are flagged (FLAG and FLAG_ROW columns)?

alex-onose commented 8 years ago

FLAG contains only zeros on all 4 columns FLAG_ROW has some non-zero entries but not on the positions where the visibilities are set to 0.

I've attached a figure vis_vs_flag_row

o-smirnov commented 8 years ago

Check also the WEIGHT and WEIGHT_SPECTRUM columns.

Also, can you put the MS and the image on Dropbox or something?

On Thu, Apr 7, 2016 at 12:35 PM, alex-onose notifications@github.com wrote:

FLAG contains only zeros on all 4 columns FLAG_ROW has some non-zero entries but not on the positions where the visibilities are set to 0.

I've attached a figure [image: vis_vs_flag_row] https://cloud.githubusercontent.com/assets/12749656/14348471/93cbbc32-fcb4-11e5-84e8-087bd49e00f7.png

— You are receiving this because you commented. Reply to this email directly or view it on GitHub https://github.com/ska-sa/pyxis/issues/78#issuecomment-206803454

alex-onose commented 8 years ago

The weights are all 1. There is no WEIGHT_SPECTRUM column.

I've put the files on dropbox

o-smirnov commented 8 years ago

OK I think we figured this one out, it was due to CASA flagging points with elevation<0.