TGBT18B_502_02 - LBW tests round two

Goals

• Test changes to shared memory setting in CODD mode
• Test that group numbers and IF tunings are consistent
• If time, do some test tone injection to double check frequency setting

Details

• PF checkout is still going at 09:30, but we'll start soon
• Made some minor script modifications to allow for different scales in different modes. Starting at 09:50
• Starting off by testing the coherent modes
• Immediately saw a bunch of IF manager errors about LOW in wrong sideband. Balance also failed.
  • There were also a bunch of errors about illegal scan number. I set it to zero instead of one. Changing that cleared a ton of errors.
  • I also had the receiver and rest frequency still set up for 820 MHz.
• Scripts are not validating after changing receiver to Rcvr_342. It validates just fine with Rcvr_800. The error seems to be in config tool. It complains with the following:

  File "/home/gbt/lib/python/telescope/gbt/config/DeviceFactory.py", line 79, in create_backend
    return VEGAS(config_table, freq_calc, ifpath, seq)

  File "/home/gbt/lib/python/telescope/gbt/config/VEGAS.py", line 51, in __init__
    self.setBanks()

  File "/home/gbt/lib/python/telescope/gbt/config/VEGAS.py", line 226, in setBanks
    self.bankSetup[bank]["mode"] = self.bankSetup["BankA"]["mode"]

KeyError: 'mode'

• • The incoherent modes seem to validate with Rcvr_342. This is specific to the coherent modes.
• Going to test incoherent modes instead. Will look at IF tuning and group number.
• Submitted SB at 10:15. Scan numbers will start with #1.
  • Complaining that balance failed. Bank E was active for some reason.
    • CM9/13 was also tuned to feed Bank E group 2 (i.e. the input IF would be 250 MHz)
  • Tried again, and Bank E was selected again.
    • vegas-hpc5's shared memory is also configured for incoherent pulsar mode.
  • Joe is very confused about Bank E. We are trying to configure with Rcvr_800 just to see what happens, even though it isn't installed.
    • Things seem to configure OK with Rcvr_800 and restfreq=800
    • Things seem to configure OK with Rcvr_800 and restfreq=350
    • Bank E activates with Rcvr_342 and restfreq=820. So it really does seem to be Rcvr_342 that is is the trigger for the issue with Bank E.
• OK, it seems like Rcvr_342 must use a different set of optical drivers, which limits it to using Rack B of the Converter Rack. Bank E is the first VEGAS bank that uses Rack B and this must be why it is being selected in config tool. This may be a hardware issue, though Steve, Galen, and Jonah aren't around to ask.
• I don't think we'll be able to test with Rcvr_342 until we figure this out. We'll try to test the shared memory set up for the coherent modes, configuring for Rcvr_800 (even though we know it isn't installed).
• Configuring for c0100x0064
  • There seems to be some issue with programming the correct BOF
  • However, looking at shared memory, OBS_BW was 100 and CHAN_BW was -1.5625, which I think is appropriate for this mode. However, OBSNCHAN was 8, and I think that should be 64. OBSFREQ is also set up as if it were getting the last 1/8 of a 100 MHz bandwidth, but it is using the correct channel bandwidth. Ray seems to understand this.
  • Got an abort during scan attempt. Tried to continue with next scan and same thing happened. Manager seems to go into Running for a little while, and then there is an abort with a "Scan terminating too early" message in CLEO.
    • Tried to cycle VEGAS Off/On and resubmitted. Same behavior - seems to go OK but then aborts.
    • It is writing a bit of data before aborting. Data do not have correct bandwidth and number of channels. This would be due to the incorrect OBSNCHAN and OBSFREQ parameters.
• We will have to call it a day. Starting version switch at 11:42
• Performed pulsar and spectral line checkouts of VEGAS and GUPPI. All looks good. Session ending at 12:10.

Conclusions

• Need to understand and work around Rcvr_342 and Converter Rack routing.
• OBS_BW and CHAN_BW have correct behavior in coherent modes but still need to fix OBSNCHAN and OBSFREQ