TGBT14A_912_87 - VEGAS bank to bank discrepancy tests
- Get spectral line data with simultaneous DCR and/or Spectrometer
Starting with Scan 9 (which is set up in a nominal configuration with low cals), we can see that we have system temperature variations, primarily in the second polarization (Note that these results are consistent when repeated consecutively, but I just show a single set of results for ease):
The value of Tcal used for calculating the Tsys is different for both polarizations, which is causes the difference in values between pol0 and pol1. These variations can also be seen in the slow switched data. Below is using scans 12 and 13 as the off and on scans:
The scaling difference are likely from using a slightly different calculation of Tsys (homemade in python for the slow switch vs GBTIDL for the normal switching). These differences are hopefully irrelevant, considering the discrepancy between banks can still be noted.
The Converter Rack output running to bank EO was then amplified and split into banks E1-H1 (EO was left disconnected). The 300 MHz high pass filter used for the noise source tests was also included. After adjusting the attenuation so that VEGAS receives a good signal (~20dBFS), slow switch data was taken. Using scan 18 as the off and scan 19 as the on gave tsys values of :
All values are consistent (at least reasonably). Similar results were found when using switched data (scans 24-26) (though you need to account for the fact that GBTIDL has no way to know it is the same signal in both pols, and thus uses the wrong Tcal value for the second polarization). The same results were found after removing the 300 MHz high pass filter (scans 30-32).
To test if it could be a linearity problem, 3, 5 and 9 dB pads were placed into E1, F1 and F2, significantly lowering the input power into those VEGAS banks. The measured Tsyss were still all consistent, suggesting the discrepancy is not a linearity issue in VEGAS. The pads were then removed.
E0, F0, G0, and H0 were then connected to the splitter (i.e. to the same converter module) and E1, F1, G1 and H1 were all connected to their normal converter rack connections for scans 39-43. These scans show acceptable deviations for both polarizations.
The final test was to connect poth polarizations of banks E and F to the splitter, and both pols of G and H to their normal converter rack connections (scans 44-49). Scan 44 shows the following Tsys:
Banks E and F use the same Tcal for both pols, and a different Tcal is used for both pols of G and H (as is done in GBTIDL). Note that you can now see the deviation in G1 and H1.
What have we learned: If VEGAS sees the same converter rack signal, it returns the same result. If VEGAS sees the same signal, but with different attenuations, it returns the same result. This suggests the problem is somehow related to how the signal from the converter rack is interpreted by VEGAS, since we get a different result from different converter modules. This does not coincide with what we have seen with the spectrometer. Grasping at straws, I have noticed that the analog filter rack uses both high and low pass filters, and the VEGAS analog system does not. Could this be a clue?
I do not quite understand the results from the last two sets of tests. Why do scans 39-43 (when all pol 0s had the same signal, but pol 1 saw their nominal converter modules) give consistent results, when scans 44-49 (when banks G and H both pols were set up nominally) give bad results in bank G and H pol 1?
Using Standard IF system
Exploring IF system with John. Configure as follows:
- Pick off Rx8_10 IF in IFrack, send to DCR 2/4. Can select filter; affects both DCR and Optical receiver, as does balance. Target level is 1, but could go up to 3. Filter can be all pass or band pass.
- Through Converter Rack, get whatever filter is set in IF rack only. Converted to baseband (~150 MHz) in Converter Rack. 750MHz should come out around 750 MHz.
- In converter Rack, we could connect one copy of receiver If to VEGAS, and one to DCR (through similar, but not identical converter rack components).
- So, could end up with:
- Rx8_10 L straight into DCR 2
- Rx8_10 R straight into DCR 4
- Rx8_10 L / R into two VEGAS banks through converter rack (4 inputs)
- Rx8_10 L / R out of above VEGAS IF modules of above banks into DCR (4 inputs)
- Rx8_10 L / R out of converter rack, into AFR, into DCR (bypasses VEGAS IF) (4 inputs)
- Output of one VEGAS IF module into power meter
Before VEGAS IF module
- Take one IF cable out of Converter Rack, into splitter, 7 splitter cables into inputs of VEGAS IF modules; eighth into Power Meter
- Outputs of VEGAS IF module into ADC, and DCR (via monitor point)
After VEGAS IF module
- Take one IF cable out of back of VEGAS IF module, 7 splitter cables into VEGAS ADC inputs, eighth into Power Meter
- Configure for TP with cal in VEGAS as if we were doing normal observing, to check that X-band looks sensible, and problem remains in the GBTIDL Tsys values
- X-band in place, boom retracted
- run bank_gain_mode4_xband (tp with cal, lo cal diode)
- balance (target is -20dBFS) balance is off by ! 1.5dB between pols in bank F, G
- Track scan 1 pol 0 Tsys ~ 45K, Tcal 2.05, pol 1 Tsys ~ 162K Tcal 1.91K
- X-band seems to be broken - switch to C-band
- bank_gain_mode4_cband (tp with cal, lo cal diode)
- balance (with noise diodes turned off)
- Track, scans 2,3,4
- bank_gain_mode4_cband_hical (to turn diodes back on)
- Track 5,6,7
- Turns out the X-band was set up for satellite mode - so we can just set the bias gains to 0dB, that fixes everything
- Return to X-band
- Track, scan 8 (not in focus)
- Track, 9,10,11
- Ok, pl 1 is back, and spread in Tsys values is around 10% (26.77, 29.27)
- bank_gain_mode4_xband_vegas_dcr_config (are VEGAS IF outputs are connected to DCR? Not sure) lbw_gains all set to be the same
- DCR banks 2 and 4 do not show 320 MHz filter in IF path, but I'm sure it is there.
- Set DCR to be it's switching signal master, VEGAS to be its.
- scan 12 noise off
- scan 13, noise lo
- scan 14, noise hi
- scan 15, noise off
- scan 16, noise lo
- scan 17, noise hi
Splitter Before VEGAS IF module
- Connect E0 (J9) Converter output to amplifier, attenuator and splitter E1 through H2
- set test attenuator to 30dB
- Set converter rack E0 to 19.5dB (for noise diode off)
- gets VEGAS dBFS to ~ -20dBFS
- Have not reconfigured or balanced anything.
- scan 18, noise off
- scan 19, noise lo
- scan 20, noise hi
- scan 21, noise off
- scan 22, noise lo
- scan 23, noise hi
- manually switch to tp mode in VEGAS and DCR, run lo cal and hi cal
- scan 24,25,26 lo cal. One pol seems higher than the other. I have been forgetting to read the power meter....
- scan 27,28,29 hi cal. VEGAS and DCR seem to have blanking enabled.
- I cannot read the power meter
- remember GBTIDL will use the wrong Tcal for for of this.
- take 300MHz high pass filter out
- scan 30,31,32 - low cal
- scan 33,34,35 - high cal
- Put 3, 6 and 9dB pads into E1, F1 and F2
- set external attenuator to 25dB * scans 36,37,38 lo cal
- connect E0,F0,G0,H0 to splitter, E1,F1,G1,H1 to regular converter rack connections
- X1 values are still balanced at -20dBFS
- ext attenuator to -30dB
- scans 39,40,41 lo cal
- scans 42,43 hi cal
- Connect E0,E1,F0,F1 to splitter, G0,G1,H0,H1 to regular converter rack connections
- Scans 44,45,46 lo cal
- Scans 47,48,49 hi cal
- 4pm Finito