Data Reduction Meeting 2018-10-04 11:00 UT

TIP Last Modified: KazushiSakamoto - 2018-10-04

Agenda and Minutes

  • [Sergio]: Observation & Data delivery Status
  • [Kazushi, Seb]: Reduction Scripts, Flux Scaling
    • [Kazushi] Script update
      • Made a ver. 3.1 script (to be posted on the ALCHEMI wiki in a couple of days)
        • manually set clean thresholds in SPWs with bright lines, where tclean overestimates rms and stops clean prematurely.
        • increase the frequency of minor cycle for some SPWs with bright lines, where clean is otherwise unstable.
        • can now apply the ver 3_1 scaling corrections to visibilities before imaging
    • Previous Flux-scaling Discussion:
    • [Kazushi] new ver. 3.1 scale factors.
      • mostly for 12m. 7m values are within ~0.2% from the previous ver. 3 values.
      • amplitudes need to be devided by these numbers
      • Band 6
        • B6a={'7m':0.957, '12m':0.986},

          B6b={'7m':1.011, '12m':0.999},

          B6c={'7m':1.015, '12m':1.027},

          B6d={'7m':1.012, '12m':1.003},

          B6e={'7m':1.005, '12m':0.984},

          B6f={'7m':0.905, '12m':0.991},

          B6g={'7m':1.059, '12m':1.003},

          B6h={'7m':0.964, '12m':0.992},

          B6i={'7m':1.000, '12m':1.006},

          B6j={'7m':1.072, '12m':1.008}

      • Band 4 - waiting for the full delivery of the 12m data
      • Band 7 - auto-masking parameters need to be optimized to avoid false masks, which may bias scale measurements.
    • [Kazushi] Test 1
      • Do the following in Band 6. 1) Obtain 7m and 12m scaling factors, independently, using the overlap methods. 2) Scale 7m and 12m visivilities using those factors. 3) Make 12m+7m cleaned cubes using the scaled data. 4) Extract spectra from the 12m+7m cubes and derive scaling factors to see how well they are aligned at the overlapping channels.
      • Result: 7m scaling factors have up to 10% deviation from unity. 12m scaling factors have up to 3% off from unity. The 12m+7m data made after rescaling have (residual) scaling factors up to 1% off from unity.
      • Conclusion: Flux alingment up to ~1% precision seems possible. (Caveat: Upper and lower halves of band 6 were observed without overlap. Relative scaling between them is not possible with the overlap method.)
      • see ngc253.scaled.12m7m.Band6.TH2.3.png for a large, 12m+7m spectrum at TH2 (the nucleus)
      • see ngc253.scaled.12m7m.Band6.S2.3.png for a large, 12m+7m spectrum at S2 (line forest)
    • [Kazushi] Test 2
      • Are there SPW-based amplitude gaps? How do the gain calibrators look?
      • Do the following for each of some Band 6 tunings. 1) Collect the pipeline flux measurements of gain calibrators (in hifa_gfluxscale) from webLog. 2) For each execution block, fit the measured flux densities with a linear function (upper panel in the figure below) and then devide each measurement with the fitted value. 3) Plot those normalized measurements (lower panel in the figure below). If there are significant SPW-to-SPW variation of scaling factors, in particular between the two SPWs in the same sideband, that should show up here as significant deviation from unity.
      • Results:
        • 7m B6a, where Seb reported 5% gap betwen the two SPWs in LSB and 3% gap between the two SPW in USB.
          • flux.B6a.7m.png
          • Mean normalized amp (bottom panel) has a 0.05% gap between the two LSB spectral windows, and a 0.7% gap btween the two USB spectral windows.
        • 7m B6b, where Seb reported 5% and 4% gaps between the two SPWs in LSB and USB, respectively.
          • flux.B6b.7m.png
          • Mean normalized amp (bottom panel) in each sideband has a < 0.1% gap between the two SPWs.
        • 12m B6a, where Seb reported 3% gap between the two SPWs in LSB, and 1% gap between the two SPWs in USB.
          • flux.B6a.12m.png
          • Mean normalized amp (bottom panel) in each sideband has a < 0.3% gap between the two SPWs.
        • 12m B6b, where Seb found 1.8% gap between the two SPWs in LSB, and the SPW with CO in USB did not work out well.
          • flux.B6b.12m.png
          • Mean normalized amp (bottom panel) in each sideband has a < 0.4% gap between the two SPWs.
      • Conclusion: SPW-to-SPW difference of scaling factors appears to be < 1% judging from the gain calibrator flux measurements checked so far. Probably it is mostly << 1% for the 7m data because several execution blocks are averaged.
      • [Seb]: Kazushi's method is a least-squared method, so the results will have uncertainties.
      • Tasks:
        • Check the gain calibrator data for all tunings, to see if there are any outliers with large scaling gaps among SPWs.
        • Are the SPW-to-SPW gaps that Seb found due to imaging (e.g., non-uniform clean depths or clean mask issues)? For fair comparison between the continuum-method and overlap-method, should test the former on the cubes cleand with the latest ver 3.1 scripts.
    • Test 3:
      • Compare 7m scaling factors with 15" beam and smaller beams (8" in Band 7 and 10" in Band 6). Compare scaling factors using spectra sampled at TH2 (dyn. center) and S2 (continuum peak with the thickest line forest).
      • Results:
        • In Band 7, 8" and 15" scale factors differ up to 3% in a couple of tunings with among the strongest lines, even though the lines are not at the overlapping channels.
          • Presumably due to some subtle imaging issues. Adopted scales from the smaller beams because what we ultimately care is 12m+7m data at < 2" resolution.
          • ACA-only paper may need re-scaling or separate scaling parameters.
        • Difference between TH2 and S2 scale factors are up to ~1%. Adopted TH2 values
        • [Seb:] Is difference between SPWs with strong lines due to dynamic range constraints? Not clear. Could test by imaging with clipping-out the strong lines.
    • [Seb:] Wants to image all data with uniform beam, but casa imaging bug is preventing this.
    • [Kazushi:] How do you control clean?
      • [Seb:] Number of iterations and RMS. Don't know which criterion stops the cleaning.
    • [Seb:] Any new news on casa mosaicing bugs? No new news.
  • [Kouichiro, Nanase]: Continuum subtraction
    • [Nanase:] No new progress.
    • [Nakanishi]: No bugs in statcont.
    • [Seb]: Taked with Alvaro last week. Statcont is not sensitive to the input RMS provided. RMS is rederived from the data itself.
    • [Nanase:] Would like to have a method which tests whether statcont is working. Check for number of channels with absorption (see last meeting notes).
    • [Jeff:] Flux scaling and continuum subtraction process will produce a well-calibrated continuum measurement that people can use to derive physical contributors to continuum (Nakanishi has a project for doing this).
  • [Stefanie]: Moment maps
    • Still working through some tweaks to the CubeLineMoment script for usability.
    • Working on convolution to same beam.
  • [Tanaka:] Investigated odd extended feature seen in CO measurements. Selfcal did not remove this feature. Will post information on how he selfcaled to wiki.
  • [All]: List of Remaining DR tasks
    • No additional tasks to discuss.
  • [All:] Will wait until after Collab meeting (next Friday) before sending around doodle poll for next DR meeting.
Topic revision: r2 - 2018-10-04, JeffMangum
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding NRAO Public Wiki? Send feedback