You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 26 Next »

2023 goals:

  1. SRDP spring release to promote new capability and use in AUDI
  2. Fall 2023 (i.e. by end of June)
    1. heuristic extensions
    2. weblog cleanup
    3. facilitate use for manual imaging scenarios that require selfcal  - gain real world testing benchmarks
  3. Fall 2024: if sufficient validation has occurred, hope to recommend for ALMA ops.

Necessary for SRDP release

  1. need regcal and selfcal aggcont and cube images created and exported PIPE-1474 - Getting issue details... STATUS
    1. filenames PIPE-1710 - Getting issue details... STATUS
    2. image headers  PIPE-1448 - Getting issue details... STATUS
  2. DR modifiers for continuum and line data split; DR modifier for continuum takes into account self-calibration improvement  PIPE-1452 - Getting issue details... STATUS
  3. Print pre/post beam info in per-solint table (weblog) -Done
  4. skip mosaics implicitly (defined in PIPE-1445) - Done
  5. 3 digits past decimal for all values (weblog) - Done
  6. Fix beam comparision ratios in per-solint stats (weblog) - Done
  7. Change in nterms for high S/N sources for hif_makeimages to nterms=2  PIPE-1772 - Getting issue details... STATUS
  8. Recipe updates for ALMA and VLA  PIPE-1559 - Getting issue details... STATUS
    1. datatype awareness for hif_editimlist?
    2. AUDI needs to still finish pipeline and make exportable images if selfcal fails
    3. unclear if forcing datacolumn='corrected' in hif_editimlist will succeed or not in cases where selfcal data are missing (if data type does not exist, shouldn't run the task)

Post-SRDP Release (for Fall 2023 i.e. by end of June)

Selfcal heuristics (RI priority order 3/21)

  • Selfcal restore  PIPEREQ-217 - Getting issue details... STATUS
  • Use parallel model writes, avoiding CAS-13338  PIPE-1789 - Getting issue details... STATUS
  • Implement low S/N and long baseline heuristics from PS (https://github.com/psheehan/auto_selfcal/tree/lbimprovements)
  • Implement mosaic heuristics from PS (https://github.com/psheehan/auto_selfcal/tree/lbimprovements%2Bmosaics%2Bcocal)
  • selfcal should reject pipeline mitigations internally
    • agg. cont self-calibration should use full image size and pixels per beam (model creation could be compromised otherwise)
    • perform self-calibration even for mitigated sources - extremely beneficial from archival point of view
  • Expose all relevant parameters
  • Determine DR correction modifications for self-calibrated line data  PIPEREQ-249 - Getting issue details... STATUS
  • Possible improvement
    • uv-plane determination if a source is detectable in field of view
    • enable inclusion/exclusion in hif_selfcal
  • Reference antenna determination cleaned up
    • VLA and ALMA use pipieline function to determine the refant list, not the function from prototype

QA & messages

  •  Fix and suppress some warnings
    • "channel bandwidths ratio?" => see  PIPE-1447 - Getting issue details... STATUS is the result of averaging down to 15.625 MHz channels when there was pre-averaging
    • suppress representative target warnings
  • main table:
    • solints to attempt - make the successful solints bolded
    • Mouseover text for success (define as: 'self-calibration solutions accepted for at least one solution interval')
    • Remove phasecenter, cellsize, and imsize from table
    • Remove the 'apply contline' and 'apply cont' columns. If those fail as part of the task, the entire task should fail.
    • Add stop reason to table

Main page and results table:

  • the page navigation with expanding sections is not used anywhere else and will be very unfamiliar - some of it is better than what is used elsewhere, but either way we should be more consistent
    • at the very least the buttons that expand alternative sections should have little "v" down arrow, or something indicating that they'll expand.
      or more consistent with the rest of the weblog would be to put all that on a subpage.
    • Some preference for this to open a new page → per 03/22 PLWG: leave as is but start with expanded + more intuitive button to go back to the top → For PS: find weblog ticket and post this decision there 1446
  • Need 'View Other QA images'-type page page to show model, residual, mask, PB, and PSF. Does not need to exactly mirror iter0, iter1 of pipeline, but should show the necessary details for each image. => Maybe not PB; make blinkable from before → after (per image type)
  • the weird 3-dot, 3-line symbol at the start of the line is that takes you back up to the top - probably just ditch that for now or have explicit ('jump to top text')
  • add mouseover text to Dist./Ratio that defines what this column is
  • theoretical sensitivity should be plotted as filled, dark gray distribution rather than the horizontal bar, this will fix many issues with the current horizontal bar
  • success "yes" looks like a clickable link but it isn't
  • down the line it would be good to figure out how to display the large table of numbers graphically - its a lot of numbers to digest.  Or as a first step highlight table cells that indicate something interesting or potentially troubling is hapenning
  • the table rendering is joining columns that probably shouldn't be joined, like "rms improvement" and "ratio of beam area" when they're both 1.0
  • need to add residual and possibly model images
  • 'N.F.' → 'Near Field' and abbreviation needs a mouse-over definition
  • I would put stop reason higher
  • General comment: Do not abbreviate words if space is available.
  • For large numbers of spws, do not list all windows. Since selfcal is done per-band for the VLA, e.g., 2~64  instead of listing all windows individually.
  • Add number of seconds in a scan

QA Plots (per-solint)subpages:

  • Need the "BACK" button like other subpages,
  • Arrow buttons to go to other solints (similar to hif_makeimages 'View other QA images' pages
  • The weird symbol with three dots and three lines doesn't seem to do anything except randomly port me somewhat upwards on the page, remove or make explicit what it does
  • Passed "Yes" again is a link that doesn't do anything when clicked
  • "Sol summary" needs better explanation
    • sol. Summary - > selfcal solution interval summary
    • 'Sol.' in row names → Solutions
  • Need 'View Other QA images'-type page page to show model, residual, mask, PB, and psf.
  • The prior/post plots should show the same information per solint that is shown on the main page for consistency.
  • Add typical scan length in seconds next to 'inf_EB' and 'inf'; e.g., inf_EB (4400s)   |    inf (180s)  |

Stop Reason reporting:

  • Make summary table
    • After selfcal task, a brief summary should be written to the CASA log for selfcal that should mirror the summary table at the top
      • Start with 'Begin Selfcal Summary...'
      • List target, band, last successful solint, stop reason (comma separated)
      • End with 'End Selfcal Summary.'
    • Something similar would be useful for the AQUA report, ALMA will need to define what they need.

Proposed QA metrics

  • S/N and RMS metrics
    • overall S/N increased, but near-field S/N decreases
    • S/N increases, but RMS increases
    • S/N decreases and RMS decreases
    • reductions in S/N for some prior/post per-spw images ( if we decide we want to be making these)
    • Look at vis amp vs. uvd for correction of decorrelation
  • Flagging metrics
    • Excessive flagging on certain EBs
      • most relevant to many EB datasets where a couple EBs with lots of failed solutions would not result in beam changing significantly
    • beam size increases (within the allowed 5%)
    • Number of long baseline antennas flagged (and passed through)
  • Calibration metrics
    • Multiple reference antenna changes on final accepted solution interval in gaincal
    • Large phase corrections on >2 antennas
      • Define large (> 90 deg?)
      • Separate determination for inf_EB and inf
        • inf_EB - system issues
        • inf - phase decorrelation



  • No labels