TGBT15A_915_99 - Detailed fftshift and scale tests for incoherent modes

Goals

• Find best combination of fftshift and vegas.scale to eliminate/minimize FFT overflow artifacts and get good output levels from VEGAS.

Details

• Switched to Justin's VPM branch
• Set iniitial fftshift to 0x5555555f and vegas.scale to 1000. Will use GUPPI when possible to verify data. Initial guppi.scale set to 1.5.
• Using IncoherentFFTSHIFTTest SB
  • SB submitted at 09:47
  • VEGAS was slow to activate and SB aborted. Resubmitting at 09:51
  • VEGAS aborted again with message "failed to start HPC subprocess". Manager reports "hpc_process: unable to open pipe: No such device or address"
  • Restarting VEGAS manager on vegas-hpc1
  • Resubmitting SB at 09:57
  • Failed again. Manager logs seem to report an issue with registering the BankAMgr. Calling Justin.
  • Justin will try to rebuild and reinstall the manager.
  • Resubmitting SB at 10:07
  • Still got an abort but trying to run scan anyway.
  • Scan #4 - No VEGAS data. GUPPI scale is high.
  • Still aborted on next scan. Maybe using GUPPI is the issue? Seems unlikely but will give it a try with just VEGAS.
  • Still failing after many attempts. Trying to run the VPM-TestManyModes SB, which ran yesterday.
  • No difference.
  • Justin reverted the manager back to the previous version (no support for 32-channel modes). This worked yesterday, so let's see if it helps.
  • Well it started running, but not in pulsar mode...aborting.
  • Got it to configure for pulsar mode, but still aborted. Put guppi_daq in MONITOR and packets are flowing, so it doesn't seem to be the ROACH.
  • Will try resetting ROACH. That is the last thing I can think of.
  • No change. Confirmed that spec line modes run.
  • Tried a different fftshift value, in case that was doing something to the ROACH. No difference.
  • Trying to do the complete switch_vegas routine again. No difference.

Conclusions