Other links:
using the 12m,fs,rdbe at AO: an overview

---

Setup:

    On 11oct21 data was taken with the 12meter, field system, rdbe backend, and mark6 recorder.
The setup was:

• 12meter tracking blank sky. The az, el was around -135, 30 deg so the telescope was looking back towards the control room..
• xband receiver used:
• udc set to center 8550 Mhz at  1500Mhz  (the center of the 1-2 Ghz IF). It is an upper side band (not flipped)
• the 1500 Mhz was converted to 750Mhz with a hi side lo (so the 750 band is flipped).
  
• 8550 is centered at 750Mhz. This is sent to the rdbe's.
• The rdbe was set to:
• use 1 rdbe: RDBE2
  
• 2 channels each 16 Mhz bw
• polA channel 1, polB channel 2.
• The 2 bit samples were then sent to the mark6 recorder in vdif format.
• 2 tests were run
• t003  - netside=u,u   . upper sideband
• t004  - netside=l,l     . lower sideband.. to see what how the parameters, outputs changed.
  
• A tone at 8552 Mhz was transmitted from the control room back to the 12 meter telescope.
• The signal was seen in the if/lo using the spectrum analyzer prior to being sent to the rdbe
• tone seen at 1502Mhz in 1-2 ghz IF (unflipped)
• tone seen at 748Mhz in the 500-1000Mhz IF (this band is flipped in freq)

The observing sequence consisted of:

• two separate runs were done
• t003 - with netside=u,u
• t004 - with netside=l,l
  
• offline:
• run sched on the T003.key file
• cp  t003.vex files to vlbis1 (online)
• repeat for t004
  
• online
• run vex2snap to create the .snp file
• start field system
• schedule= start running the schedule
• adjust the power levels in if/lo, a/d in rdbe and then digitizer quantization.
• let rdbe take data for 60 seconds.
  

Files used for the experiment:

• 	generated by sched			generated by vex2snap
  exp	.key	.sum	.vex	.snp	.prc
  t003	t003.key	t003.sum	t003.vex	t003aa.snp	t003aa.prc
  t004	t004.key	t004.sum	t004.vex	t004aa.snp	t004aa.prc

---

Taking the data

• start the field system (on vlbis1)
  
• fs
• schedule=t003aa   (or t004aa)
• # wait for rdbe to initialize
• # set the attenuator for the a/d (outside of field system)
• /usr2/oper/bin/
• RDBE2_AGC_AR.py  (for the single rdbe we are using)
• # Response:  result = !dbe_rms=0:19.839:18.895:-0.810:-0.588:127:84;
  
• a/d 1 sigma rms was close to 20 counts
&$
• # set the quantization levels
• ddc_quantize_read_adj_set.py
  
• # this does all ddc's , all rdbe's
• # it will first try to adjust the a/d again.
• # there is lots of output. the final quantize query for each rdbe/ddc looks like:
• Response:  result = !dbe_ddc_quantize?0:0:5119992:10881625:10875224:5123159:1959:-1960:0;
• The ratio of the levels was: .47,1,1,.47
• For each test 60 seconds of data was taken.
  

---

Looking at the data

• logged at oper@mk6 to get the data.
  
• The raw files were combined with vsb_fs. the combined files were:
  
• ~oper/data/tmp/t003_aa_no0001/t003_aa_no0001
• ~oper/data/tmp/t004_aa_no0001/t004_aa_no0001
• The combined (but  not vmuxed) files were copied to gpuserv0:/media/DATA0/phildata/vlbi/211011/  and then processed using some idl routines i've written (to input, demux, and print the headers)l
• I also ran vmux, m5spec.py to verify that i got the same results at the idl routines.
• vmux  t003_aa_no0001/t003_aa_no0001 5032 1600 0,1 t003_aa_no0001.vmux
• (and similar for t004)
• arguments:
• samples/frame= 5000bytes*(8bits/2bits) = 20000
• sec/frame=1/32MHz * 20000 =.000625seconds
• frames/sec=1/.000625=1600 frames/sec/chan
• m5spec.py t003_aa_no0001.vmux VDIF_10000_128_2_2 1000 16384
• and similar for t004
• arguments
• 10000 bytes data/frame. since we put 2 5000 byte frames in 1 frame
• 128Megasmp/sec: 32MhzSampling*2bits*2chan=128.
• 1000: integrate for 1000 millisecs
• 16384 to a 16K xform -> 976.6 hz channel resolution.

Print the headers from the combined (but not vmuxed file)

• 20 headers from t003_aa (.txt)
• 20 headers from t004_aa (.txt)

There is a problem with the frames from both experiments:

• FrameNum Epoch  Seconds  Frame Thread Length Chans Bits L I C EDV SampleRate   SyncWord DBE IF Sub Tuning(MHz) Side Rev Pers
        37    43  8883000     37      0   5032     1    2 0 0 0   3   16000000 0xacabfeed   1  0   1   10.00000    L  1.6 0x83
        37    43  8883000     37      1   5032     1    2 0 0 0   3   16000000 0xacabfeed   1  1   1   10.00000    L  1.6 0x83
        37    43  8883000     37      0   5032     1    2 0 0 0   3  128000000 0xacabfeed   1  0   2    0.00000    U  1.6 0x83
        37    43  8883000     37      0   5032     1    2 0 0 0   3  128000000 0xacabfeed   1  0   2    0.00000    U  1.6 0x83
• There are 4 frames for each time stamp rather than 2.
• thread 0, samplerate 1600000, tuning 10Mhz is polA
• thread 1, samplerate 1600000, tuning 10Mhz is polB
• thread 0, samplerate 128000000, tuning 0  . is probably the 3rd output of the rdbe with default parameter settings
• thread 0, samplerate  128000000,tuning 0  . is probably the 4th output of the rdbe with default parameter settings.
  
• --> It looks like the rdbe is outputting 4 threads rather than 2.. even though we only setup the 1st 2.

The voltage distribution and the spectra.

    In idl

• 4096 frames of voltage data were input and demultiplexed into the two pols.
  
• the incorrect frames : samplerate =128000000 were excluded
• This was done twice for each experiment:
• using output thresholds of : [-3.3349,-1,+1,3.3340]  to compute the spectra
• using output thresholds of : [0,1,2,3]    when making the histograms

The plots show the spectra and voltage histograms for the 2 experiments:

• t003   spectra/histograms (.ps) (.pdf)
• This used netside=u,u in the .key file
• page 1: spectra
  
• The top plot is polA, the bottom frame is polb
• the setup parameters are listed
• The  tone (at 8552 Mhz) is showing up at + 2MHz
• The center of the band: 8Mhz should be 8550 Mhz, so the tone should be showing up at 8+2Mhz.
• It really is an upper sideband. On a previous day we transmitted 8553 Mhz and it showed up at +3 Mhz.
• i computed the spectra via
• do an  fft on 16384 real samples at a time (the imaginary part is 0.)
• this gives a 16384 complex output where the spectra is 8k long and then repeated in the 2nd 8k
• I used the 1st 8k .  abs(fft)^2 , integrated spectra and then plotted .
• Page 2: voltage histograms:
• The 4096 frames (using lut values [0,1,2,3] for min to max values were used to make a histogram
• Black in polA, red is PolB,
  
• Top: counts for the 4 channels
• Bottom: normalize the spectra to the average of the -1,+1 counts
• The blue plot is the histogram taken from the dbe_quantize command.
• The ratio of values in the histogram is close to .4. the expected value should be .47
• The voltage rms was computed:  1sigma=.915 and .907 digitizer levels. The optimum values for floating point would be .995. Integer values should give values close to .94.
• So the quantization levels should have been a bit closer to 0. (to increase the outer level counts,
• t004_spectra/histogram (.ps) (.pdf)
  
• this used netside=l,l in the .key file
• i wasn't clear on whether netside was me telling sched what i wanted, or me telling sched what the sideband really was.
• The tone did not appear in the spectra.
• the voltage histogram was a little closer to the optimum value. ratio about ..47
• rms measured: .95. rms optimum (integer)=.94..
  

---

The tone does not appear in the expected freq channel

    I'll step through the tone locations in the different bands:

• Analog:
•  1-2 GHz IF (unflipped)
  
• 1500 Mhz = 8550Mhz
• 1502 Mhz  = 8552 tone. this was verified on the spectrum analyzer
• 500-1000 Mhz IF (flipped)
• 750 Mhz = 8550 Mhz
• 748 Mhz = 8552 Mhz tone. this was verified on the spectrum analyzer
• This band was sampled by the rdbe a/d
• digital
• The pfb ranges below are just my guesses
  
• 
  

• 	pfb#	Rf  Center range	IF center Mhz	RF center (8550)	tone (8552)
  a/d input (1024Mhz clk)	-	8532 8788-8276	768	750	748
  a/d output (0 - 512 Mhz)	-	8532 8276-8788	256	274	276
  pfb 128 mhz channel inp src @0 to 512 mhz after digitizer flip	0	8340 8276-8404	64 0 to 128	-	-
  	1	8468 8404-8532	192 128 to 256	-	-
  	3	8596 8532-8660	320 256 to 320	274	276
  	2	8728 8660-8788	448 320 to 512	-	-
  pfb 128 mhz chan outputs	0	8340 8276-8404	0 -64 to 64	-	-
  	1	8468 8404-8532	0 -64 to 64	-	-
  	3	8596 8532-8660	0 -64 to 64	-46	-44
  	2	8728 8660-8788	0 -64 to 64	-	-

•  
• The header showed that the mixer was set to 26Mhz (not the  -46) i have above.
• If the pfb band was  0 to +128Mhz then
• 0 hz would  be 8532 Mhz
• + 18Hz would be 8550 Mhz
• +26Mhz  would be bw/2 above the edge of the 16 Mhz band
• The tone would be at +2 Mhz.
• So this matches what was seen in the output. If the specified freq is the edge of the band and not the center.
  

---

Summary

• It looks like the rdbe is always outputting 4 bands (even if you only setup 2)
• Requesting a freq of 8550 Mhz and 16 Mhz bw:
• a 8552 Mhz tone showed up at 2 Mhz rather than 8+2= 10 Mhz
• The netside=u,u seemed to put the tone close to where it should be.
• the final output band should have been upper sideband (from the 750Mhz flipped band, and the flip from the undersampling).
• netside=l,l .. chose a different  pfb channel and the tone was not seen.
  

processing:x101/211011/fstest.pro