TGBT18B_502_09 - Additional LBW tests

Goals

  • Look for correct frequency labels in Modes 1--3
  • Investigate switching signal control in i0100x1024 mode
  • If time, look at fftshift and scale value for HBW modes

Details

  • Joe and Ray started switching over at 11:35
  • First scans submitted at 12:00. Will test frequency labels in Modes 1--3, total power only. Ray and Joe think they fixed this and that it was ultimately an issue with the units being used.
    • All these look good now!
  • Will do a spot check of Modes 4, 7, 10, 20, and 29 to make sure they still look good in total power.
    • Modes 4, 7, and 10 looked OK
    • Scan aborted on Mode 20. Bank D aborted due to scan terminating too early
    • Resubmitted starting with Mode 10
      • Remaining scans ran OK and look good
  • Will do a spot check of Modes 1, 4, 7, 10, 20, and 29 with frequency switching
    • There are huge baseline ripples in Modes 4 -- 20, and lots of narrow birdies (in groups) visible in Modes 10 and 29. Not sure what these is from.
  • We are going to switch back to the release branch quickly to see if these structures are still there.
  • Resubmitted frequency switching scans at 12:35
    • Forgot to take Antenna out of scan coordinator, so first scan aborted. First scan in this sequence is #18
    • Ripples are still there on release branch. Birdies are not but those could have been transient (it is maintenance day). It could have be one of the standing wave ripples. At any rate, it doesn't seem like it is any worse on the new branch.
  • Switched back to LBW branch and just taking two more scans in Modes 10 and 29 to see if birdies diminished.
    • Online SDFITS was down, so good scans are #26 and 27
    • Birdies are gone. It could easily have been something related to maintenance activities
  • Spectral line modes look good
  • Will now investigate switching signal in i0100x1024 mode
    • LED on SSDS seems to be flashing at correct rate
    • Joe says that the Blanking light is also on. Not sure if that is legitimate or not.
    • Tried a 0200x1024 scan to check Blanking state
      • Looks like it is active there, too, but cal is seen.
    • Going to check 0100x0512 to look at bandpass levels
      • Levels look similar
    • Very confusing. Bandpass and switching states all look the same but still no cal in i0100x1024. Going to run another scan to look at noise diode states in CLEO for PF
      • Nothing looks off by eye
    • Going to try a scan with high power diode
      • CLEO indicates that the diode is in the correct state. ADC input power also seems to be varying by about 3 dB in both polarizations, as we might expect. But still no cal.
    • Going to try a search-mode scan with a 1.0 second switching period to see if we can fold and find it.
      • Found some clues. In the i0100x0512 mode (which works) and the above set up we see the diode blinking at 1 Hz as expected. The scan length was 60 seconds and the number of samples recorded reflects this.
      • In i0100x1024 mode (which doesn't work) the diode is there, but it appears as if it is switching at 1.5 Hz (0.6666 seconds, 2/3 the requested value) and the scan length as calculated from the number of samples is 40 seconds, 2/3 of expected.
      • Randy will investigate this.
    • We also hooked a scope up to the switching signal distribution system and verified that the signal is at 1.0 seconds
  • Ryan had to go to an UWB meeting at 14:15
    • Back at 14:45
  • Will now look at some different fftshift values for the i1500 modes and the variances we get in the bandpass
    • Taking scans in i1500x[512,1024,2048] modes with fftshift = 55555557
      • vegas.scale probably needs to be lowered by about half in 512 and 2048 channel mode. Might be OK in 1024 channel mode.
    • Double cal signal in 2048 channel mode!
    • Mean and variance looks good in Banks A and B, decent in C, variance is a little low in D
    • Decreasing vegas.scale to 5000 from 10000 in 512 and 2048 channel mode
    • Going to decrease fftshift to aaaaaaab
      • Not a significant difference in variance
      • vegas.scale is OK in 512 and 2048 channel mode. Lowering to 3000 from 10000 in 1024 channel mode/
        • 3000 looks good
    • Going to try a different frequency set up per bank
      • Low frequencies (high baseband frequencies) are suppressed a bit. Difficult to get good balance on Bank C
    • Going to get a baseline scan with the original fftshift value of aaaaaaaf, fftshift set back to 10000
      • These start with scan #52. Power levels were too low on Bank A
    • Going to try L-Band to look for FFT artifacts with different fftshift values.
      • fftshift = aaaaaaab start with scan #55.
      • Levels were too high and I actually wanted to do 800 MHz modes. Starting over with scan #56. Setting vegas.scale to 5000
        • Overflow issues don't seem to bad in 512, 1024 channel modes. Some issues in 2048 and 4096 channel modes but not too bad.
      • Going to try fftshift of 55555557 and vegas.scale of 5000. First scan is #60.
        • fftshift didn't seem too bad except for a bit in 2048 channel mode
  • Ray and Joe started switching back to release version at 16:00
  • Checkouts look good. Session ended at 16:20

Conclusions

  • Frequency labels look good in Modes 1--3
  • Frequency switching looks good in Modes 1, 4, 7, 10, 20, and 29
  • We found some clues about switching signal in i0100x1024 mode, but Randy needs to look BOF file
  • I think we can move to a smaller fftshift value without too much risk of overflow at L-Band
-- RyanLynch - 2019-02-07
Topic revision: r2 - 2019-02-07, 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