Intro:

The setup:

• tx setup:
• dual beam with 800 kw to dome , 350 kw to ch
• dome moving .04 deg/sec between 5 and 19 degrees (continual).
• az=270
• iflo setup.
• 430 -> 750 -> 260 -> 260Amp -> mixer (260 to 30) -> 20Mhz bw 30Mhz filter -> rdev in other room
  
• @ 260 the IF is unflipped
• 260-> 30 mixer we found in the drawer.
• Used 290 lo (7dbm) so 30 Mhz IF is flipped
  
• had rms volts about 50 millivolts. for data
  
• Rdev setup :
  
• image:/pcmn/sasdr_image333.gz
• 140 MHz clock.
• config file:pnet.conf.fir35M_20131108
• tx samples a/d 1
• rx samples a/d 2
  
• dec_f=1 35Mhz bw
  
• tune: 28087,14043H = 920350135 = fcenter*2^32/140 = 30Mhz
  
• txSamples=17500
• s2_count=21000 .. start height window at 600 usecs or 90 km
  
• hghtSamples: 129500 .. covers 3.7 millisecs or 555 km.
• shift_fir = 8 . upshift value for fir output.
  
• sps:
  
• running 10 milliseond ipp, 440 usec clp pulse.
  
• Software:
• test run on pdevs0.
• scripts in ~thall/
• conf=pnet.conf.fir35M_20131108
• img=/pcmn/sasdr_image333
• pnet --conf=$conf --fpga=$img  --extclk --dump=100000
• data was stored in .pdev format with 20 gb file sizes.
• 
  

The data files recorded:

    The following datafiles were taken:

• All files under /share/pdata0/pdev/
• You should access the files using /net/pdevs0/pdev/  
  
• some files are 20gb. The /net mounts use the newer version of nfs that works with large files

• basename	size gb	startTime ast hh:mm:ss	nfiles
  sasdr_world.20131108.b0a.00100.pdev	2	12:24;24	1
  sasdr_world.20131108.b0a.00200.pdev sasdr_world.20131108.b0a.00201.pdev .. sasdr_world.20131108.b0a.00205.pdev	20 each file	12:25:52	6

The tx and rx windows:

Tx window before adding 2 samples/ipp

• Data was taken from the first 100 ipps (1 sec of data) of file sasdr_world.20131108.b0a.00100.pdev.
• Top: overplot the I voltage samples for the first 10 ipps. (the first 400 samples of each ipp are shown).
• The first 60 usecs blanked by the iflo blanking.
  
• The transitions are not occurring at the same spot
• Bottom: start of tx pulse for the first 100 ipps. over plot with offset for display.
• You can see the 1st blanking region drifting to the right for each ipp
• The start end of the drift are flagged : samples 38 and 237.
  
• the tx pulse is moving by 200samples/100 ipps or 2 samples/ipp
• The samples that come in on the left look like noise samples (from the heights).

Tx window after adding 2 samples/ipp to the reported values.

• Page 1: start of tx pulse
• frame 1,2: the i, q voltage samples for the start of the txpulse
• 100 ipps are overplotted
• bottom: blowup showing a single tx ipp.
  
• each + is a sample. I is blank Q is red
• The dashed green lines are spaced by 1 baud (2usecs or 70 samples).
• Page 2: end of tx pulse
• 100 ipps are overplotted
• Top is i, bottom is Q voltages
• Adding 2 samples to each ipp lines up the tx pulses.

The receive window:

• Page 1: voltage samples start of receive window.
• Top i voltages
• Bottom: q voltages
• 25 ipps have been overplotted with an offset for display.
• the first 45 usecs of the receive window has no signal (from the blanking).
• You can see each ipp moving to the left by about 4 samples.
  
• After 8 ipps the window jumps by about 32 samples back to the starting value
• The dashed vertical lines is the farthest right the blanking region attained,.
• Page 2: power
• Top: compute power for each sample then smooth.
• 300 meters of data were smoothed  for each ipp, then 100 ipps were averaged.
• Bottom: average bandpass for receive window
• The spectral density was computed (using a 64K xform) for the start of each rx window (skipping the blanked region).
• The spectra where then averaged for 1 second, converted to db and then plotted.
• The shoulder at 422 Mhz is the 20 Mhz filter (at rf)  after the lna.
• The bump in the center is probably the spectrum of the code from the ion line.

• You can see the end of the blanking moving and then coming back every 8 ipps
• The dashed lines show the range of the drift (30 samples)
  
• My guess is that my lines are a little off. It is probably 4 samples/ipp or 32 samples in 8 ipps.

10 second decoded dynamic spectra:

    Dynamic were made using 10 sec blocks of decode data. The images came from file sasdr_world.20131108.b0a.00200.pdev.

• 2 extra samples were read for each ipp  to align the data
• The receive window shift was not corrected.
• 32k transforms were done to compute the spectra.
• spectra were kept every baud length (2 usecs).
• It took 15 minutes (in idl) to process 1 10 sec block.
• Each image has the freq axis smoothed by 25 bins, and heights by 2 bins.
• The last image is a blowup where the freq is only smoothed by 6 bins.
  

Summary:

• rdev is recording 2 extra samples for each ipp (using the tx and rx window lengths as the reference).
• This was corrected in the offline software by increasing the samples/ipp by 2.
  
• the receive window is moving by 4 samples/ipp to the left (it is missing samples).
  
• After 8 ipps it jumps back to where it started.
• So the receive window is moving  in time by 32 samples  (or about 1 usecond) every 80 millisecs)
• This was not corrected in the offline software.
• Dynamic spectra show the upshifted and down shifted plasma line.
• I probably need to spend some more time figuring out the optimum median filter length to remove the noise.
• We only recorded 1630 heights. I spaced them by 2 useconds (or1 baud). We probably want to double this.
• The i/o rate from spectrometer to file server (about 65 Mbytes/sec) is limiting the number of heights.
• We can trade off bandwidth channels for more heights.
• Changing the clock frequency will let us tune this (as long as the fpga doesn't mind)

<- page up
home_~phil