TGBT15A_915_103 - More fftshift and scale tests for incoherent modes

Goals

  • I previously tried to find good values of vegas.scale for an fftshift value of 0x5555555f. The vegas.scale values were uncomfortably high, so I will try to an fftshift values of 0x55555557 and see if lower vegas.scale values will work. Will also test some new BOF files from Luke:
    • i0800x1024_x14_7_24t_095binw_2017_Jun_22_1648.bof
    • i1500x1024_x14_7_24t_095binw_2017_Jun_22_1020.bof
    • c1500x1024_x14_7_24t_095binw_2017_Jun_26_0950.bof
    • c0800x2048_x14_7_24t_095binw_2017_Jul_11_1123.bof
    • c1500x2048_x14_7_24t_095binw_2017_Jul_17_1526.bof

Details

  • Session begins at 09:22
  • Switched to Justin's VPM branch
  • Using IncoherentFFTSHIFTTests SB
  • Will use vegas.scale = 1000 as initial value for all modes
  • SB submitted at 09:41
    • Scan #1 is i0800x0032
      • Levels are much too low
    • Scan #2 is i0800x0064
      • Levels are pretty much perfect
      • No apparent overflow issues
    • Scan #3 is i0800x0128
      • Levels are good
      • No apparent overflow issues
    • Scan #4 is i0800x0256
      • Levels are good
      • No apparent overflow issues
    • Scan #5 is i0800x0512
      • Levels are good
      • No apparent overflow issues
    • Scan #6 is i0800x1024
      • Levels are a little low, by maybe a factor of 5-6
      • Maybe some overflow artifacts? Tough to tell
    • Scan #7 is i0800x2048
      • Levels are much too high
      • There seem to be overflow artifacts but they are only really visible in the parts of the band with no signal
    • Scan #8 is i0800x4096
      • Levels are a little low, by maybe a factor of 5-6
      • Overflow artifacts don't seem to be an issue
    • Scan #9 is i1500x0032
      • Levels are good
      • No apparent overflow artifacts
    • Scan #10 is i1500x0064
      • Levels are good
    • Scan #11 is i1500x0128
      • Levels are high by a factor of about 3
      • No apparent overflow artifacts
    • Scan #12 is i1500x0256 * Levels are OK but could lower by about a factor of 1.5 * No apparent overflow artifacts
    • Scan #13 is i1500x0512 * Levels are are high by a factor of about 3 * No apparent overflow artifacts
    • Scan #14 is i1500x1024 * Levels are low by about a factor of 2-3 * Possibly seeing some minor overflow artifacts
    • Scan #15 is i1500x2048 * Levels are good * Possibly seeing some overflow issues
    • Scan #16 is i1500x4096 * No data
  • Will try again with same fftshift value. Some scales have been changed: i0800x0032 --> 5000, i0800x1024 --> 5000, i0800x2048 --> 200, i0800x4096 --> 5000, i1500x0128 --> 300, i1500x0256 --> 700, i1500x0512 --> 300, i1500x1024 --> 2500
  • SB submitted at 10:34
    • Scan #17 is i0800x0032
      • Levels look good
      • No apparent overflow artifacts
    • Scan #18 is i0800x0064
      • Levels look good
      • No apparent overflow artifacts
    • Scan #19 is i0800x0128
      • Levels look good
      • No apparent overflow artifacts
    • Scan #20 is i0800x0256
      • Levels look good
      • No apparent overflow artifacts
    • Scan #21 is i0800x0512
      • Levels look good
      • No apparent overflow artifacts
    • Scan #22 is i0800x1024
      • Levels look good
      • Starting to see possible overflow artifacts
    • Scan #23 is i0800x2048
      • Levels look good
      • Seeing overflow artifacts
    • Scan #24 is i0800x4096
      • Levels look good
      • Possibly seeing overflow artifacts
    • Scan #25 is i1500x0032
      • Levels look good
      • No apparent overflow artifacts
    • Scan #26 is i1500x0064
      • Levels look good
      • No apparent overflow artifacts
    • Scan #27 is i1500x0128
      • Levels look good
      • No apparent overflow artifacts
    • Scan #28 is i1500x0256
      • Levels look good
      • No apparent overflow artifacts
    • Scan #29 is i1500x0512
      • Levels look good
      • Seeing some overflow artifacts
    • Scan #30 is i1500x1024
      • Levels look good
      • No major overflow artifacts
    • Scan #31 is i1500x2048
      • Levels look good
      • Overflow artifacts are starting to appear in low-power parts of band
    • Scan #32 is i1500x4096
      • Levels are OK but could be a bit higher
  • I'm worried about overflow issues in some of the modes. Will set fftshift to 0xaaaaaaaf and reset all vegas.scale to 1000
  • SB submitted at 11:15
    • Scan #33 is i0800x0032
      • Levels are too low
    • Scan #34 is i0800x0064
      • Levels are a little low but reasonable
      • No apparent overflow artifacts
    • Scan #35 is i0800x0128
      • Levels are low by a factor of 2-3
      • No apparent overflow artifacts
    • Scan #36 is i0800x0256
      • Levels are a little low but reasonable
      • Maybe some minor overflow artifacts?
    • Scan #37 is i0800x0512
      • Levels are low by a factor of 2-3
      • No apparent overflow artifacts
    • Scan #38 is i0800x1024
      • Levels are low by a factor of 5-6
      • No obvious overflow artifacts
    • Scan #39 is i0800x2048
      • Levels are way too high by about a factor of 3-5
      • Possible overflow artifacts in low-power part of band
    • Scan #40 is i0800x4096
      • Levels are low by a factor of 5-6
      • No apparent overflow artifacts
    • Scan #41 is i1500x0032
      • Levels are low by a factor of 5
      • No apparent overflow artifacts
    • Scan #42 is i1500x0064
      • Levels are good
      • No apparent overflow artifacts
    • Scan #43 is i1500x0128
      • Levels are good
      • No apparent overflow artifacts
    • Scan #44 is i1500x0256
      • Levels are good
      • No apparent overflow artifacts
    • Scan #45 is i1500x0512
      • Levels are OK, could maybe be higher by a factor of 1.5
      • Maybe a slight hint of overflow artifacts
    • Scan #46 is i1500x1024
      • Levels could be higher by a factor of 2
      • No apparent overflow artifacts
    • Scan #47 is i1500x2048
      • Levels are too low by a factor of 10
      • No apparent overflow artifacts
    • Scan #48 is i1500x4096
      • Levels are too low by a factor of 2-3
    • Scan #49 IS i0800x0032
      • Levels are still too low
    • Scan #50 is i0800x0064
      • Levels are good
    • Scan #51 is i0800x0128
      • Levels are good
    • Scan #52 is i0800x0256
      • Levels are good
    • Scan #53 is i0800x0512
      • Levels are good
    • Scan #54 is i0800x1024
      • Levels are good
    • Scan #55 is i0800x2048
      • Levels are much too high
    • Scan #56 is i0800x4096
      • Levels are good
    • Scan #57 is i1500x0032
      • No data
    • Scan #58 is i1500x0064
      • Levels are good
    • Scan #59 is i1500x0128
      • Levels are good
    • Scan #60 is i1500x0256
      • Levels are good
    • Scan #61 is i1500x0512
      • Levels are
    • Scan #62 is i1500x1024
      • Levels are good
    • Scan #63 is i1500x2048
      • Levels could be lower by a factor of 1.5
    • Scan #64 is i1500x4096
      • Levels are good
    • Scan #65 is i0800x0032
      • Levels are still low by a factor of about 2
    • Scan #66 is i0800x2048
      • Levels are still too high
    • Scan #67 is i1500x0032
      • Levels could be a bit higher
    • Scan #68 is i1500x2048
      • Levels are good
    • Scan #69 is i0800x0032
      • Levels are still low
    • Scan #70 is i0800x2048
      • No longer saturated by still high
    • Scan #71 is i1500x0032
      • Levels look good
    • Scan #72 is i0800x2048
      • Now a bit too low. Going to vegas.scale = 500
    • Scan #73 is i0800x2048 with vegas.scale = 500
      • Still a bit high.
    • Scan #74 is i0800x2048 with vegas.scale = 300
      • Close enough
  • Switching back to release branch at 14:28
  • Submitted CheckMode1 and CheckMode29 SBs
    • All looks good

Conclusions

  • While there seem to be some OK combinations of fftshift and vegas.scale, I'm concerned that some modes still either need too large a vegas.scale parameter or seem to be showing some overflow artifacts. WIll need to think about better ways to mitigate this.
-- RyanLynch - 2017-07-19
Topic revision: r2 - 2017-07-20, 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