VEGAS Bank Discrepancy Problem - 13 January 2015

More attempts to diagnose the cause of this problem, after the requantization balancing fix.

Experimental setup

Data was taken into TGBT14A_912_60. Here is the observing log.

The telescope was parked at access, and VEGAS was configured with the L-band receiver, and frequency switching, so that the HI line is present in the passband. VEGAS was configured to use all eight banks, and in the high modes, all eight subbands. The rest frequency in every case (8 or 64) was set to 1420.0. So, apart from using different IF paths, VEGAS should have been seeing the exact same signal. In all cases, the input and requantization output power levels were spot-checked to ensure they agreed with the target values. This was true in all cases (to within the 2dB tolerance) apart for a one-off scan (scan 6).

As a comparison, the GBT Spectrometer was configured into 800 MHz, four bank mode, and the same observing procedure used.

Tsys results

The Tsys values as reported by the GBTIDL "getfs" procedure are listed in the table below. Bank H was observed to be consistently more discrepant than the other banks (see results in the observing log), and so the mean and standard deviation values were calculated with and without Bank H.

Modes not in this table have not been tested in this way.

scan mode bof pol meanTsysAll stddevTsysAll meanTsys7 stddevTsys7
2 3 H16K/HBW 0 17.6952 0.176311 17.6564 0.149002
2 3 H16K/HBW 1 18.3252 0.185047 18.3047 0.189772
3 3 H16K/HBW 0 17.3661 0.125499 17.3375 0.103669
3 3 H16K/HBW 1 18.3057 0.137986 18.2903 0.141465
4 acs - 0 15.9451 0.655466 - -
4 acs - 1 16.4584 0.174451 - -
5 acs - 0 15.2856 0.618735 - -
5 acs - 1 15.9165 0.089848 - -
6 21 L8/LBW1 0 16.6754 1.99006 16.2983 1.81472
6 21 L8/LBW1 1 16.5289 1.72128 16.5832 1.85178
7 21 L8/LBW1 0 17.9058 2.14904 17.4309 1.81189
7 21 L8/LBW1 1 17.9109 2.11800 17.9714 2.28020
11 21 L8/LBW1 0 100.052 10.9559 97.6135 9.19398
11 21 L8/LBW1 1 98.0930 11.8315 98.2095 12.7745
12 21 L8/LBW1 0 16.5269 1.82128 16.1190 1.52231
12 21 L8/LBW1 1 16.8366 2.08225 16.8726 2.24640
13 21 L8/LBW1 0 16.8890 1.91608 16.4392 1.54758
13 21 L8/LBW1 1 17.1144 2.11979 17.1488 2.28722
14 11 L8/LBW1 0 16.5614 0.23613 16.5322 0.23895
14 11 L8/LBW1 1 17.1625 0.25673 17.1779 0.27328
15 11 L8/LBW1 0 16.2498 0.25910 16.2158 0.25988
15 11 L8/LBW1 1 16.9087 0.23712 16.9181 0.25450
16 4 L1/LBW1 0 16.1410 2.29668 15.3782 0.85099
16 4 L1/LBW1 1 16.7300 0.13332 16.7126 0.13389
17 4 L1/LBW1 0 16.0743 1.89399 15.4719 0.89345
17 4 L1/LBW1 1 16.9145 0.13556 16.9146 0.14642
18 4 L1/LBW1 0 16.0665 1.80118 15.4735 0.70880
18 4 L1/LBW1 1 16.8933 0.14790 16.8935 0.15975
20 11 L8/LBW1 0 16.4941 0.26227 16.4663 0.27020
20 11 L8/LBW1 1 17.0307 0.26306 17.0429 0.28169
21 21 L8/LBW1 0 17.1489 2.02901 16.6768 1.65020
21 21 L8/LBW1 1 16.9800 1.81205 16.9977 1.95649
22 7 L1/LBW1 0 16.6945 1.81124 16.0572 0.19049
22 7 L1/LBW1 1 17.0463 0.24220 17.0530 0.26081
24 24 L8/LBW1 0 18.4235 2.19879 17.8527 1.61243
24 24 L8/LBW1 1 18.4122 2.35513 18.4502 2.54117
25 21 L8/LBW1 0 17.9767 2.16702 17.3542 1.36440
25 21 L8/LBW1 1 17.9544 2.54027 17.9916 2.74145
27 21 L8/LBW1 0 15.41498 1.740096 14.93110 1.160724
27 21 L8/LBW1 1 16.09989 2.260974 16.15782 2.435710
29 21 L8/LBW1 0 16.50400 2.082292 15.90493 1.307287
29 21 L8/LBW1 1 18.60376 2.681180 18.66548 2.889861
Notes:
  • In scan 6, Bank D did not balance properly.
  • For Scan 11. the boom was deployed
  • Scan 12 was done with the input power level, and requantization power level set to -26dBm (6 dB lower than normal), and Scan 13 to -14dBm (6dB higher than normal).
  • Scan 25 was done wiith an eight second switch period, and 64 seconds scan duration.
  • For scans 27/28, and 29/30, the Cal On and Cal Off scans were performed separately as individual tp_nocal scans, and the system temperature calculated manually, using the Tcal value from the GBTIDL header.

Discussion

  • For the ACS, the bank-to-bank variation is around 4% for pol 0, and about 1% for pol 1. Previous observations have shown variations at about the 1% level.
  • For Mode 3, both polarizations show variations at the < 1% level.
  • For Modes 4 and 7, pol 0 consistently shows variations at the 12% level, while pol 1 shows variations at the 1% level.
  • Mode 11 consistently shows variations at the ~ 1.5% level in both polarizations
  • Mode 21 shows variations in both polarizations at the > 10% level. This is consistently true whether the input power level is changed by +/- 6dB, the Tsys value is changed by a factor of ~6, or the integration time is increased.
  • Mode 24 shows variations in both polarizations at the >10% level. Mode 24 is the eight subband analog of Mode 11.
  • Mode 21 shows variations at the xx % level in yy polarizations, even when the observations are done with no switching and no blanking (and hence completely identical integration times).

Conclusions

  • The effect is not present in Mode 3 (H16K/HBW).
  • The effect is present in one polarization for two L1/LBW1 modes (4 and 7) but not the other. * The effect is not present in an L8/LBW1 single-subband mode (mode 11), but it is there in the exact analog eight subband mode (mode 24).
  • The effect demonstrated in mode 21 is present independent of switching, blanking and integration time, and is constant over a wide range of Tsys and input power level.
  • What the heck is going on?
  • Where do we go from here?
  • Investigate mode 4/7 pol 0 / 1 - e.g. by switching polarizations? Capturing raw packets somewhere, and processing them independently?
  • Try Anish's suggestion of splitting the output from one analog module, and feeding it to many banks (so the signal is "exactly" the same)? I predict this will fail also.
Topic revision: r2 - 2015-03-07, RichardPrestage
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