09/03/2023 Meeting Notes

[Victoria-Samboco]

Today we discussed and analysed the results of the Suns subtraction Workflow which I explain below:

1. Self-calibration of the main field(Required)
In this step we proceed with the calibration and imaging of the observation

2. backup_UVW (Not Required)
Backup of the UVW coordinates before modifying it

3. restore_UVW (Not Required)
Restoring the UVW coordinates after modifying it, when needed.

4. scan_numbers(Required)
A script to extract a list of unique scan numbers from a MS file. The scans output parameter is a list of strings representing the unique scan numbers.

5. split_ms_by_scan(Required)
A step to split the MS into scans using as input the ms and the scan_numbers

6. backup_UVW_scans (Not Required)
Making a backup of the UVW coordinates of the splitted MSs before modifying it

7. get_original_coords(Required)
A step that gets the original phase centre coordinates of the splitted MSs from the PHASE_DIR column to use later in the phase rephase_ms_to_old_phase_centers to rephase back to the original phase centre. In this step we get the phase coordinates in radians from the PHASE_DIR column and convert it to hms and dms and write it in a txt file called old_coods.txt

8. sun_coordinates (Required)
Determine the Sun coordinates (RA/DEC) for all scans in the MS and write in a txt file, this is then used to change the phase centre of the MS's to point to the Sun.

9. shift_to_sun (Required)
Change the phase centre of the scans using the outfile of the sun_coordinates step.

10. image_sun (Required)
Making a dirty image of the Sun for a list of scans.

11. create_ds9_regions (Required)
This step finds the position of the Sun written in a txt file (sun_coordinates.txt) in ra and dec in the format hms and dms , and create a circular ds9 region file centred on each position (line) with a diameter of 0.5 degrees.

12. make_mask (Required)
create masks for each scan using the ds9 region files created in the step below for a list of MSs

13. update_ms_files (Necessary but Not required)
This step use the taql command to add columns (corrected_data+model-data) when needed

**14. clean_image_sun-Deconvoluton of the Sun ** (Required)
Deconvolution of the Sun with Wsclean. In this step we try to find the best model for the Sun to use it for the Subtraction step

15. rename_model_data (Required)
Before predict the model of the Sun we use this step which uses the talql command to rename the MODEL data to ORIGINAL_MODEL_DATA , as Wsclean will automaticaly take this column to perform subtraction.

16. dd_predict (Required)
Predicting the model of the SUN, After predicting I will have the model of the Sun in the column MODEL_DATA

17. subtract_sun (control step)
Subtract the model of the Sun (MODEL_DATA) from the rephased CORRECTED_DATA.

18. restore_UVW_scans (Not really Required because to get the RA/DEC of the original field I use the PHASE_DIR column)
Restore the UVW from the backup to the UVW column, for a list of scans

19. rephase_ms_to_old_phase_centers (Required)
Rephase the MSs scans to the old phase centre using the old_coods.txt file as input.

20. subtract_sun_from_vis (Required)
Subtract the model of the Sun (MODEL_DATA) from the original visibilities (CORRECTED_DATA)

21. quartical_2 (Not required yet)
Self calibration in the main field with then Sun Subtracted

22. image_5
Imaging after the main field proceeding with Wsclean with the Sun subtracted.

Then the same steps are done but now for the scans splitted by timeranges.

This are the images of the main field before and after Sun subtration ( the second image is the dirty image with no cleaning (niter=0)

Its possible to see from the image that the the emission from the aliased sunspot decrease,meaning that we had some subtraction, but not enough yet.

During the meeting was suggested (by @IanHeywood ) to use the padding: 2 to rather than 1.5 that was being used, to produce higher quality images with improved sensitivity and with the hope that it will remove the aliasing.

Question: I Might also put the recipes together with the explanation?

[IanHeywood]

Thanks for the summary @Victoria-Samboco.

• I'm a bit confused about the "restore UVW" step. Doesn't chgcentre deal with this?

• Depending on how QuartiCal is configured, you might need to have a step to setup an appropriate model column between steps 20 and 21. This could simply be making sure QuartiCal is using ORIGINAL_MODEL_DATA, but alternatively you could do a masked run of wsclean following the subtraction of the Sun to setup a deep model. As I said in the meeting if the solar interference is bad enough then the sidelobes that are washing over the target field will corrupt the initial model.

• It was @landmanbester who made the suggestion of using padding: 2. :)

On that note, I think you deconvolve the Sun-subtracted data and increase the padding to remove the aliases we should be able to have a proper before and after comparison and see how well your process has worked. I'm a bit confused as to the before and after images above though, I'm sure the ones you showed on your screen share earlier looked like the sidelobes from the Sun were much less pronounced... I think upping our bookkeeping game was another point that emerged from the meeting!

Great work, thanks.

[Victoria-Samboco]

Thanks for the summary @Victoria-Samboco.

• I'm a bit confused about the "restore UVW" step. Doesn't chgcentre deal with this?

Yes that was from the first workflow, where we wanted to save the UVWs to use later to rephase back to the visibilities, but now we don't need it anymore.

• Depending on how QuartiCal is configured, you might need to have a step to setup an appropriate model column between steps 20 and 21. This could simply be making sure QuartiCal is using ORIGINAL_MODEL_DATA, but alternatively you could do a masked run of wsclean following the subtraction of the Sun to setup a deep model. As I said in the meeting if the solar interference is bad enough then the sidelobes that are washing over the target field will corrupt the initial model.

Okay, in fact I've already a script that rename back the columns I just need to now incorporate it in the pipeline

• It was @landmanbester who made the suggestion of using padding: 2. :)

On that note, I think you deconvolve the Sun-subtracted data and increase the padding to remove the aliases we should be able to have a proper before and after comparison and see how well your process has worked. I'm a bit confused as to the before and after images above though, I'm sure the ones you showed on your screen share earlier looked like the sidelobes from the Sun were much less pronounced... I think upping our bookkeeping game was another point that emerged from the meeting!

Yes, its a good point, I will incorporate it in the pipeline also.

Great work, thanks.

Thank you :)