TGBT15A_915_79 - Test New PFB Core for c0800x0512 Pulsar Mode

Goals

Ryan discovered artifacts in coherent VEGAS pulsar mode data taken as part of astrophysical acceptance testing. Specifically, there is extra noise in certain frequency channels, which seems to repeat in structure and amplitude after some number of frequency channels, when inspecting pulsar phase vs frequency plots. This does not appear to be affecting pulsar fluxes but it is obvious in observations of the noise diode. When discussing this Randy, Ryan learned that the coherent mode BOF files had been built with a sub-optimal PFB core---the PFB bin width was larger than that used in GUPPI, which should result in a few dB less suppression of out of channel power. Per Randy, during GUPPI commissioning Paul Demorest noted artifacts that were related to the PFB bin width. Though it is not clear if the artifacts seen in VEGAS are the same, there is no doubt that the coherent mode PFB cores should use a smaller bin width, and this may also get rid of the artifacts we currently are seeing. Randy built a new BOF file for the c0800x0512 mode using a 12-tap PFB with 0.95 bin width (identical to GUPPI) for doing a quick spot check. The plan for today is to test this new BOF file along with the old, sub-optimal version, to check for basic functionality and to determine if the artifacts noted above are absent. The new BOF file is c0800x0512_095bws_x14_7_2017_Jan_20_0730.bof

Details

Session begins at 10:47
  • Switched VEGAS to pulsar branch
  • Performing 5 pulsar cal scans with the original BOF file (c0800x0512_x14_7_2016_Mar_14_1421.bof)
    • Using AcceptanceTestingCODDPrep scheduling block
    • SB submitted at 10:49
    • First scan is #1, last scan is #5
    • Artifacts are visible
    • SB ended at 10:58
  • Switching to new BOF file (c0800x0512_095bws_x14_7_2017_Jan_20_0730.bof)
    • Using AcceptanceTestingCODDPrep scheduling block
    • SB submitted at 11:01
    • Scan aborted on first attempt (failed to start the VEGAS HPC subprocess)
      • Ray is trying to run guppi_daq_server manually but is getting a bus error and core dump.
      • Ray asks Ryan to turn the manager off/on and resubmit the SB. Ryan does so.
      • guppi_daq_server did not start after turning managers back on.
      • Ryan tries doing a full VEGAS system stop/start
      • vegas_hpc_server seems to have started after system came back up in mode 1
      • Astrid froze, but Ryan restarted and resubmitted SB
      • guppi_daq_server seems to have started OK
    • SB resubmitted at 11:27
    • Scan aborted on resubmit
      • VEGAS HPC program taking too long to be ready on vegas-hpc3 (bank B)
      • Ryan edited vegas.conf to use original BOF file and turned manager off/on. guppi_daq_server did not seem to restart.
      • Ryan tries doing a full system stop/start again
    • SB resubmitted at 11:41
      • Took one scan, which ran OK
  • Switch back to new BOF file
    • Using AcceptanceTestingCODDPrep scheduling block
    • Resubmitted SB at 11:47
    • First scan is running OK with new BOF
    • First scan is #10, last scan is #15
    • Artifacts still seen, but spot check of data looks otherwise OK
    • SB ended at approx 12:00
  • Switching back to release branch
    • Had to do a second restart of turtle
  • Running 1 spectral line scan for verification
    • Using CheckMode1 scheduling block
    • Submitted SB at 12:08
    • Balanced failed; EXCEPTION: can't convert power value to float
    • Balanced successfully on second attempt
    • Scan is #15
    • SB ended at 12:09
    • SD fits was not running; Grail and turtle restarted
    • Resubmitting SB at 12:12
    • Scan is #16
    • SB ended at 12:14
    • Spot check of scan looks OK in gbtidl
  • Running 1 spectral line scan in mode 22 for verification
    • Using CheckMode22 scheduling block
    • Submitted SB at 12:14
    • Scan is #17
    • SB ended at 12:15
    • Accidentally had receiver = 'NoiseSource 'in config
    • Resubmitting SB at 12:17 with NoiseSource = 'Rcvr2_3'
    • Scan is #18
    • SB ended at 12:17
    • Spot check of scan looks OK in gbtidl
  • Session ending at 12:18

Conclusions

There are still manager issues, but the new BOF file seemed to perform OK. However, it did not get rid of the data artifacts.

-- RyanLynch - 2017-01-18
Topic revision: r2 - 2017-01-25, RyanLynch
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