TGBT15A_915_98 - Test Improved BOF Files

Goals

  • Test new 32-channel BOF files and BOF files for other bandwidth and mode combinations with 24-tap, 0.95-bin-width PFB components. BOF files are
    • i0800x0032_x14_7_24t_095binw_2017_Jun_13_1343.bof
    • i0800x2048_x14_7_24t_095binw_2017_Jun_19_1655.bof
    • i0800x4096_x14_7_24t_095binw_2017_Jun_19_1155.bof
    • i1500x0032_x14_7_24t_095binw_2017_Jun_22_0831.bof
    • i1500x2048_x14_7_24t_095binw_2017_May_15_1540.bof
    • i1500x4096_x14_7_24t_095binw_2017_Jun_08_1610.bof
    • c0800x0032_x14_7_24t_095binw_2017_May_25_0716.bof
    • c0800x2048_x14_7_24t_095binw_2017_Jun_20_1600.bof
    • c0800x4096_x14_7_24t_095binw_2017_May_18_1157.bof
    • c1500x0032_x14_7_24t_095binw_2017_May_24_0720.bof
    • c1500x2048_x14_7_24t_095binw_2017_Jun_20_1205.bof
    • c1500x4096_x14_7_24t_095binw_2017_May_24_1042.bof

Details

Session begins at 10:00
  • Switched to Justin's VPM branch using 'switch_vegas pulsar'. Will test i0800x2048, i0800x4096, c0800x2048, and c1500x4096 to start.
  • Will use VPM-TestManyModes SB.
    • SB submitted at 10:07
    • Scan #1 - i0800x2048
    • Scan successful. Levels are low but noise diode is visible.
    • Scan #2 - i0800x4096
    • Scan successful. Levels are low again but noise diode is visible.
    • Scan #3 - i1500x2048
    • Scan successful. Levels are low again but noise diode is visible.
    • Scan #4 - i1500x4096
    • About 7% dropped packets. Nonetheless, levels are low again but noise diode is visible.
    • Scan #5 - c0800x2048
    • Data were not recorded on banks C, D, E, and H, but there were no error messages from VEGAS, either. Diode is visible on banks with data.
    • Scan #6 - c0800x4096
    • Scan successful. Diode is visible.
    • Scan #7 - c1500x2048
    • About 80-90% dropped packets. Scan executed successfully but insufficient data to check noise diode.
    • Scan #8 - c1500x4096
    • Scan aborted with many messages about faults in roach_state, master_slave_sel, dest_port, dest_ip, scale_p1, acale_p0, and nchan
    • Cycled VEGAS on/off and tried to resubmit c1500x4096 scan.
    • Scan #9 - c1500x4096
    • Scan aborted again.
  • Problems with c1500x4096 mode, but other modes seemed OK. I changed fftshift to be 0x55555557 for incoherent modes and 0xffffffff for coherent modes. vegas.scale is 1000 for incoherent modes and 150 for coherent modes. Will redo scans for the modes that worked with this scaling to see if that improves data quality.
  • Using VPM-TestManyModes SB
    • SB submitted at 10:40
    • First scan aborted with same errors as with the c1500x4096 scan. Maybe something wrong with the ROACH? Not sure if I did an off/on cycle, though. Will try again.
    • Cycled VEGAS off/on and resubmitted SB at 10:43.
    • Worked this time. Scan #10 - i0800x2048
    • No errors but also no data.
    • Scan 11 - i0800x4096
    • Data look OK after removing RFI, except for a feature coherent in pulse phase around a phase of 0.6
    • Scan 12 - i1500x2048
    • Data look OK after removing RFI.
    • Scan 13 - i1500x4096
    • About 10% dropped packets again. Data look otherwise OK after removing RFI.
    • Scan 14 - c0800x2048
    • Lots of artifacts in bandpass. Bandpass shape seems totally wrong.
    • Scan 15 - c0800x4096
    • Data look OK
    • Scan 16 - c1500x2048
    • Lots of dropped packets again. Difficult to judge.
  • Levels were low in the incoherent modes. Will try again with vegas.scale = 5000
  • Using VPM-TestManyModes SB
    • SB submitted at 11:04
    • Scan 17 - i0800x2048
    • Bandpass levels look better. Data looks OK.
    • Scan 18 - i0800x4096
    • Bandpass levels look better. Data looks OK. No artifact in pulse phase.
    • Scan 19 - i1500x2048
    • Levels are now too high. No errors but no data written. Noticed that vegas_status died, though, with message TypeError: must be string without null bytes, not str, so possible shared memory error?
    • Scan 20 - i1500x4096
    • Levels look good. Data looks good after RFI removal.
  • Justin made some alterations to config tool to support 32-channel modes. Will restart turtle and try these.
  • Using VPM-TestManyModes SB
    • SB submitted at 11:20
    • Scan 21 - i0800x0032
    • Manager is complaining about an illegal value in nchan parameter. Justin will need to fix this.
  • Will try coherent modes again with vegas.scale set to 200.
  • Using VPM-TestManyModes SB
    • SB submitted at 11:36
    • Scan 21 - c0800x2048
    • Scan 22 - c0800x4096
    • Scan 23 - c1500x4096
    • > 90% dropped packets, but no issues with illegal parameters. I guess the Manager got stuck in a weird state last time.
  • Switching back to release VEGAS at 11:45
  • Verification scan in mode 1 looks good.

Conclusions

32-channel modes could not be tested because they are not supported by the manager yet. The other BOF files looked OK with the possible exception of c0800x2048. The caveat is that packet loss in the c1500x2048 and c1500x4096 modes makes a detailed check difficult. -- RyanLynch - 2017-03-29
Topic revision: r2 - 2017-06-28, 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