VEGAS Testing August 27 2014 Part 2 - Richard and Adam
Overview
- Attempt to better understand the noise spikes in VEGAS data
- Data in TGBT14A _912_24
Brief Background
Numerous spikes can be seen in the VEGAS data. These spikes occupy a single channel or pair of channels, and look a lot like RFI, but in bands not common for RFI. The results here utilize Q-band (Rcvr40_52), where RFI should be minimal. The spikes can more easily be seen by convolving the data with a high pass filter. In GBTIDL:
result=convol(getdata(),digital_filter( 0.1, 1, 200, 64))
It was originally thought that these spikes were limited to modes 2/3, but were recently found in
mode 4. (here, the strong spikes are at 8192 and 24576, which are multiples of 512, but there are also spikes at 13474 and 25121, which are not.) Digitally filtering the data showed significantly more spikes than initially obvious as can be seen
here. To see how consistent these spikes were across frequency space, we used three banks looking at three different frequencies (no overlapping wavelengths), all in mode 4.
This plot shows the locations of the largest spikes (defined as >1.e8) in bin-space, with spikes from each bank getting their own value (1-3). This shows that the spikes more or less appear in the same bins in every bank of the spectrometer. The lack of a 1-1 correlation is likely caused by the very rudimentary thresholding used to determine the locations of the spikes, causing some of the spikes to not be detected (as most are suppressed when comparing to
this).
Preliminary Analysis
We observed in modes 1,2,3, total power with cal, 10 minute tracks, repeated twice for each mode. Input signals were the external noise source (Bank A), and Q-band looking at blank sky, for the other banks. A typical mode 2 total-power (with cal) scan is show here:
As can be seen, no spikes are obviously visible in the ten-minute integration. However, following Ron Maddalena's suggestion, we then used GBTIDL "gettp",
and convolved the data with a digital filter. The filter response is shown here:
The results for modes 1,2,3, with the two successive scans overlaid, are shown below:
As can be seen, no spikes can be seen in mode 1, but a rail of spikes can be seen in modes 2 and 3. The spikes overlay exactly in channel
between successive scans in the same mode, but do no overlay in frequency space between modes. The spikes are quite repeatable in
height from scan to scan, but not exactly so. The spikes are not a perfect rail (i.e. not evenly spaced in frequency). Text files containing
the location of mode 2 and 3 spikes are here:
Previous data show similar spikes in mode 4, but these seem more intermittent, and were not visible in this dataset. But, spikes were seen in
higher numbered modes (data analysis still underway). Here is an example of spikes in a previous Mode 4 dataset:
To see if the spikes could be due to noise calibration, we looked for spikes in data without noise calibration:
It appears that the spikes are not dependent on the calibration switching.
Time Series Analysis
We took a 20 min scan in mode 2 with VEGAS (scan 79 in TGBT14A_912_26). At the same time, we used the archivist to record the ADCPWR1 values, (the manager grabs 16k samplers in a chunk). Adam took an FFT of each chunk separately, and then averaged the 1200 rows together, see below:
- Time series processing; raw bandpass spectrum:
- Time series processing; filtered bandpass spectrum:
We took the contemporaneous VEGAS FITS file, get GETTP, binned the data by a factor of two to get the same number (8K) of channels, and plotted the results, see below:
Here are some plots in channel space:
- In both Modes 2 and 3, spikes appear at ... 5120, 5632, 6144 ... so appears to be every 512 channels.
- The spikes have been present since at least 27 February 2014 (TGBT13B_502_37)
Varying fftshift
Joe has done some tests with the SRBS, see
https://safe.nrao.edu/wiki/bin/view/GB/Software/VegasSpikeSRBSTesting
Details
- Using SpikeTest to look for spikes in different modes:
vegasConfig = """
receiver = 'Rcvr40_52'
beam = 'B12'
obstype = 'Spectroscopy'
backend = 'VEGAS'
vegas.subband = 1
nchan="low"
deltafreq = 0
dopplertrackfreq=42000
bandwidth = 1500
swmode = 'tp'
swtype = 'none'
swper = 1.0
swfreq = 0.0
tint = 1.0
vframe = 'topo'
vdef = 'Radio'
noisecal = 'lo'
pol = 'Circular'
restfreq = [{"restfreq":42000, "bank":"A","beam":"1","bandwidth":1500}, # noise source, mode 1
{"restfreq":42000,"bank":"B","beam":"1","bandwidth":1500}, # IF, mode 1
{"restfreq":42000,"bank":"C","beam":"1","bandwidth":1500,"nchan":"high"}, # mode 2
{"restfreq":42000,"bank":"D","beam":"1","bandwidth":1080,"nchan":"low"}, # mode 3
{"restfreq":42000,"bank":"E","beam":"2","bandwidth":187.5,"nchan":"low"}, # mode 4
{"restfreq":42000,"bank":"F","beam":"2","bandwidth":187.5,"nchan":"medium"}, # mode 5
{"restfreq":42000,"bank":"G","beam":"2","bandwidth":187.5,"nchan":"high"}, # mode 6
{"restfreq":42000,"bank":"H","beam":"2","bandwidth":100,"nchan":"low"}] # mode 7
"""
source=GetCurrentLocation("J2000")
off = Offset("J2000", 0.0, 0.0, cosv=False)
Configure(vegasConfig)
Balance()
Break("Check Balance")
#SetValues(NoiseSource:noiseFilter)
#SetValues("IFRack",{'noise_bandwidth':'broadband'})
#SetValues("IFRack.IFNoise", {'noise_bandwidth':'broadband', 'state':'Ready'})
#Break("get prepared to switch in the noise source")
Track(source, None, 600.0)
#OffOn(source,off,60.)
- Scan 1: Spiketest 1 sec integration times, Modes 1,1,2,3,4,5,6,7
- Scan 2: SpikeTest 30 sec integration times, modes as above
- Scan 3: Spiketest2 30 sec integration times, modes 1,8,9,10,11,12,13,15 (SpikeTest2)
- Scan 4: as above
- Scan 5: SpikeTest3 30 sec integration times, modes 1,16,17,18,20,21,22,23
- Scan 6: as above
- Scan 7: Spiketest4 30 sec integration times, modes 1,24,25,26,27,28,29
- Scan 8: as above
- Scan 9: SpikeTestMode2_3TpNoCal
- Scan 10: as above
- Scan 11: SpikeTestMode2EightBanks
- Scan 12: as above
- Scan 13: SpikeTestMode3EightBanks
- Scan 14: as above
- Scan 15: SpikeTestMode4EightBanks
- Scan 16: as above