search

ACES-CMZ / reduction_ACES

Reduction scripts and tools for ACES
https://worldwidetelescope.org/webclient/?wtml=https://data.rc.ufl.edu/pub/adamginsburg/ACES/mosaics/mosaics.wtml
15 stars 12 forks source link

Execution Block ID uid://A001/X15a0/Xea Sgr_A_st_m_03_7M [SgrB2] #15

Open keflavich opened 2 years ago

keflavich commented 2 years ago

Sgr_A_st_m_03_7M uid://A001/X15a0/Xea

Product Links:

Reprocessed Product Links:

djeff1887 commented 2 years ago

hif_makeimages (cube)

The pipeline didn't continue past this step.

spw16 and spw18

Both of these windows have deep negative residuals over the sources: image image

I believe there's SiO in spw18, H13CO+ and HN13C may be here too but they're very narrow compared to some of the other EBs: image

spw20

There's still some structure in the residual image: image

HCO+: image

spw22

HNCO: The emission is cut off at the low frequency edge of the spectral window uid___A001_X15a0_Xea s38_0 Sgr_A_star_sci spw22 cube I iter1 image spectrum

It says tclean reached the niter limit for this window, as well: Warning! tclean reached niter limit of 499999 for Sgr_A_star / spw22 / iter1 !

spw24

There's structure in the residual here as well, I don't think it did a very good job: image

Not sure what the brightest line is here, but it doesn't have a very normal-looking shape: image

spw26

This clean was stopped to prevent divergence: Warning! tclean stopped to prevent divergence (stop code 5). Field: Sgr_A_star SPW: 26 iter1 image

The line-free moment0 shows negative values overlapping with where the sources should be, based on the residuals: image image

keflavich commented 2 years ago

Looks like just that one channel needs reimaging: image

djeff1887 commented 2 years ago

Continuum

Broadly, I think deeper/better cleaning can be done for most of these images, as shown previously there's structure in several of the residual images. The same can be said for the continuum: image image

Emailed help desk about the missing findcont images: manual continuum frequency ranges were selected to improve detecting line emission, hence why the weblog mentions an "OLD" continuum and the ranges are displayed in the imaging task spectral plots. If we'd like to see the original pipeline-determined continuum, we just have to rerun the pipeline without passing the manually-defined cont.dat file while it's running. They also pointed us to the QA2 report, since it mentions the manual continuum selection.

Some line image/identification:

H13CN

Seems to be in self-absorption here, also HC15N as the bright peak at low frequencies image

H13CO/SiO

Looked at this again, H13CO+ is the leftmost bright peak. It's narrow and relatively less pronounced because it's essentially only localized to the hot cores. SiO is the large peak near the middle of the window. Pretty sure HN13C is the narrow peak on the right, its seems to behave similarly to H13CO+. image

HC3N

The 11-10 transition is the bright line here; it's also coincident with the diverging channel in this spw. I am testing reimaging parameters as of writing this, will update with the results. image

djeff1887 commented 2 years ago

Attempted a simple reclean last week by dropping the niter parameter from 899999 to 700000 which fixed the diverging channel. For reference, I followed the divergence example from the Dan's wiki post, first making a dirty image using: tclean(vis=['../uid___A002_Xf396d6_X4ff6.ms.split.cal', '../uid___A002_Xf396d6_X54f0.ms.split.cal', '../uid___A002_Xf396d6_X6084.ms.split.cal', '../uid___A002_Xf396d6_Xa339.ms.split.cal', '../uid___A002_Xf396d6_Xa983.ms.split.cal', '../uid___A002_Xf3e416_X105aa.ms.split.cal', '../uid___A002_Xf49cca_Xdcd0.ms.split.cal'], field='Sgr_A_star', spw=['26', '26', '26', '26', '26', '26', '26'], antenna=['0,1,2,3,4,5,6,7,8&', '0,1,2,3,4,5,6,7,8&', '0,1,2,3,4,5,6,7,8&', '0,1,2,3,4,5,6,7,8,9&', '0,1,2,3,4,5,6,7,8,9&', '0,1,2,3,4,5,6,7,8,9&', '0,1,2,3,4,5,6,7,8,9&'], scan=['7,10,13,16,19', '7,10,13,16,19', '7,10,13,16,19', '7,10,13,16,19', '6,9,12,15,18', '7,10,13,16,19', '7,10,13,16,19'], intent='OBSERVE_TARGET#ON_SOURCE', datacolumn='corrected', imagename='uid___A001_X15a0_Xea.s38_0.Sgr_A_star_sci.spw26.cube.I.iter1.reclean', imsize=[288, 288], cell=['2arcsec'], phasecenter='ICRS 17:47:12.7761 -028.20.06.419', stokes='I', specmode='cube', nchan=3827, start='99.5725753248GHz', width='0.4882796MHz', outframe='LSRK', perchanweightdensity=True, gridder='mosaic', mosweight=True, usepointing=False, pblimit=0.2, deconvolver='hogbom', restoration=False, restoringbeam='common', pbcor=False, weighting='briggsbwtaper', robust=0.5, npixels=0, niter=0, threshold='0.0mJy', nsigma=0.0, interactive=0, usemask='auto-multithresh', sidelobethreshold=1.25, noisethreshold=5.0, lownoisethreshold=2.0, negativethreshold=0.0, minbeamfrac=0.1, growiterations=75, dogrowprune=True, minpercentchange=1.0, fastnoise=False, savemodel='none', parallel=True)

I then reimaged the spw using: tclean(vis=['../uid___A002_Xf396d6_X4ff6.ms.split.cal', '../uid___A002_Xf396d6_X54f0.ms.split.cal', '../uid___A002_Xf396d6_X6084.ms.split.cal', '../uid___A002_Xf396d6_Xa339.ms.split.cal', '../uid___A002_Xf396d6_Xa983.ms.split.cal', '../uid___A002_Xf3e416_X105aa.ms.split.cal', '../uid___A002_Xf49cca_Xdcd0.ms.split.cal'], field='Sgr_A_star', spw=['26', '26', '26', '26', '26', '26', '26'], antenna=['0,1,2,3,4,5,6,7,8&', '0,1,2,3,4,5,6,7,8&', '0,1,2,3,4,5,6,7,8&', '0,1,2,3,4,5,6,7,8,9&', '0,1,2,3,4,5,6,7,8,9&', '0,1,2,3,4,5,6,7,8,9&', '0,1,2,3,4,5,6,7,8,9&'], scan=['7,10,13,16,19', '7,10,13,16,19', '7,10,13,16,19', '7,10,13,16,19', '6,9,12,15,18', '7,10,13,16,19', '7,10,13,16,19'], intent='OBSERVE_TARGET#ON_SOURCE', datacolumn='data', imagename='uid___A001_X15a0_Xea.s38_0.Sgr_A_star_sci.spw26.cube.I.iter1.reclean', imsize=[288, 288], cell=['2arcsec'], phasecenter='ICRS 17:47:12.7761 -028.20.06.419', stokes='I', specmode='cube', nchan=3827, start='99.5725753248GHz', width='0.4882796MHz', outframe='LSRK', perchanweightdensity=True, gridder='mosaic', mosweight=True, usepointing=False, pblimit=0.2, deconvolver='hogbom', restoration=True, restoringbeam='common', pbcor=True, weighting='briggsbwtaper', robust=0.5, npixels=0, niter=700000, threshold='0.453Jy', nsigma=0.0, interactive=0, usemask='auto-multithresh', sidelobethreshold=1.25, noisethreshold=5.0, lownoisethreshold=2.0, negativethreshold=0.0, minbeamfrac=0.1, growiterations=75, dogrowprune=True, minpercentchange=1.0, fastnoise=False, restart=True, savemodel='none', calcres=False, calcpsf=False, parallel=True)

I am currently testing niter=899999 again, and will report back with the results. The residuals left a lot to be desired in the niter=700000 run, so I'm hoping to find a combination of parameters that can let us clean better and avoid the divergence (assuming niter=899999 is indeed causing divergence).

keflavich commented 2 years ago

@djeff1887 any updates on this? Could you add your new parameters to the override_tclean_parameters.json

ashleythomasbarnes commented 2 years ago

H13CN cleaning - try cycleniter within tclean task

djeff1887 commented 2 years ago

Testing cycleniter

After many experiments, simply changing the cycleniter from the default value to 100 seems to have fixed everything (thanks for the suggestion @ashleythomasbarnes). All other tclean parameters were identical to their pipeline counterparts: image

Not sure if everything will break upon trying to clean deeper, but that's a fight for another day. Here's the tclean command I ran, for reference: 

tclean(vis=['../uid___A002_Xf396d6_X4ff6.ms.split.cal',
       '../uid___A002_Xf396d6_X54f0.ms.split.cal',
       '../uid___A002_Xf396d6_X6084.ms.split.cal',
       '../uid___A002_Xf396d6_Xa339.ms.split.cal',
       '../uid___A002_Xf396d6_Xa983.ms.split.cal',
       '../uid___A002_Xf3e416_X105aa.ms.split.cal',
       '../uid___A002_Xf49cca_Xdcd0.ms.split.cal'], field='Sgr_A_star', spw=['26',
       '26', '26', '26', '26', '26', '26'], antenna=['0,1,2,3,4,5,6,7,8&',
       '0,1,2,3,4,5,6,7,8&', '0,1,2,3,4,5,6,7,8&', '0,1,2,3,4,5,6,7,8,9&',
       '0,1,2,3,4,5,6,7,8,9&', '0,1,2,3,4,5,6,7,8,9&', '0,1,2,3,4,5,6,7,8,9&'],
       scan=['7,10,13,16,19', '7,10,13,16,19', '7,10,13,16,19', '7,10,13,16,19',
       '6,9,12,15,18', '7,10,13,16,19', '7,10,13,16,19'],
       intent='OBSERVE_TARGET#ON_SOURCE', datacolumn='data',
       imagename='uid___A001_X15a0_Xea.s38_0.Sgr_A_star_sci.spw26.cube.I.iter1.cyclefactor2_7sigma',
       imsize=[288, 288], cell=['2arcsec'], phasecenter='ICRS 17:47:12.7761 -028.20.06.419', stokes='I', specmode='cube', nchan=3827,
       start='99.5725753248GHz', width='0.4882796MHz', outframe='LSRK',
       perchanweightdensity=True, gridder='mosaic', mosweight=True,
       usepointing=False, pblimit=0.2, deconvolver='hogbom', restoration=True,
       restoringbeam='common', pbcor=True, weighting='briggsbwtaper',
       robust=0.5, npixels=0, niter=899999, threshold='0.453Jy', nsigma=0.0, cyclefactor=1.0,
       interactive=0, usemask='auto-multithresh', sidelobethreshold=1.25,
       noisethreshold=5.0, lownoisethreshold=2.0, negativethreshold=0.0,
       minbeamfrac=0.1, growiterations=75, dogrowprune=True,
       minpercentchange=1.0, fastnoise=False, restart=True, savemodel='none', calcres=False, calcpsf=False, parallel=True)
djeff1887 commented 2 years ago

Have discovered an issue with the pipeline products for spw22. The weblog reports that it reached the niter limit, and examining in CARTA shows this odd flashing issue across the (cont sub'd) image. spw24glitching2

In detail, the spectra in affected areas show "ringing," but I'm not immediately sure how to interpret it: image

The model and residual images also show the same behavior.

ashleythomasbarnes commented 2 years ago

Hmm yeah I've come across this is some other data I've been working with. Could you overlay the mask as contours and flick through the cube? In my experience, this has happened because the automasking has a strict mask in one channel, and then a broad in the next, and then a strict mask again etc etc

keflavich commented 2 years ago

this has happened because the automasking has a strict mask in one channel, and then a broad in the next, and then a strict mask again

That's exactly what happened in this case. But I've seen other cases where this happens without a mask at all also.