14feb12 puppi first light with lband

Setup:

• lbw recieiver in linear mode
• Tracking J2017+0603.
  
• IF to puppi : 1000 to 1600 (1180 to 1780 Mhz). 1200 Mhz IF centered at 1380 Mhz.
• 800 Mhz search mode using second nyquist band.
  
• 2048 channels, spectral integration 40.96 usecs.
• 53.7 seconds of data (1 11gb file) was processed:
• 4096 (40.96usec) spectra/row in fits file
• 1 row=.167 seconds
• 71 rows in file

plots and images.

• all data: Avg spectra, rms/mean by channel (.ps) (.pdf)
• Page 1: average specgtrum (53.7 seconds)
  
• Top: Stokes I
• Bottom:PolA(Black), polB (red)
• The horizontal axis is sky frequency.
• The falloff at 1120 Mhz is the bandpass filter for lband wide.
• There might be a 31 Mhz Ripple in the spectra.
• If this is a cable reflection then:
  
• 1./31Mhz * 150 Meter/usec * .7(velInCable) = 3.4 meters or about 11 feet.. So probably not the cable from the receiver room to puppi (although it probably wouldn't hurt to check it).
  
• :Page 2: rms/mean by frequency channel.
• For each row (4096 spectra) the rms is computed along each channel and then divided by the mean value (Tsys).
• Top: rms/mean for each row then average the 71 rms/means for the file
• The horizontal red dashed  line is the expected rms/mean from the radiometer equation.
• The vertical dashed lines show various rfi sources:
• Black:  punta salinas radar in mode A  (more  info)
• Red remy radar (1270,1290) (more  info)
• green: FAA radar(1330, 1350)  (more  info)
• blue Iridium satellite (more info)
• Light brown: band edges from lbw bandpass filter.
• rms/mean for each row over plotted (with offsets) for 12 seconds of data.
• You can see the punta salinas radar (1230,1250)  blank its transmitter when it points at us (3-5 seconds)
• There are a few rows who have spike all across the band (6.5 light blue, 8 secs red). This is probably the digitizer or butterflys of the polyphase filter bank overflowing.
• Dynamic spectra for  row 38 (6.4 seconds) showing radar pulses (.gif)
• This  shows the dynamic spectra for the 4096 spectra in row 38 of the 1st file (6.4 lite blue trace that showed spikes across the entire band).
  
• It covers 167 milliseconds.
• horizontal lines  of power start appearing at 20 milliseconds and last till about 120 milliseconds.
• The vertical pulses near 1620 are iridium satellite. The spacing on the lower, and middle pulses is 90 milliseconds. This is iridiums basic cycle (45 milliseconds uplink,45 millisecs downlink).
• Row 38 total power vs time showing radar pulses (.ps) (.pdf)
• Top: Total power vs time for the 4096 spectra in row 38 (lite blue line with spikes in rms/mean plots).
• You see the radar beam sweep through arecibo during its 12 second rotation period.
• The individual spikes are the radar Ipp (Inter Pulse Period of  2.5 to 3.5 milliseconds).
• Bottom: The time difference between adjacent radar spikes (ipps).
• The FAA radar uses 5 ipps to get around range aliasing:
  
• 2.633,2.821,2.74596,2.594.84,3.330996 milliseconds
• The colored dashed lines are the 5 ipps used by the FAA radar.
• The Lineup shows that these total power  spikes (or horizontal lines in the spectra) are coming from the 1330,1350 faa airport radar.
• I'm a little surprised that there isn't a larger spike in the spectra around 1330, 1350. It may just be the way the image is scaled.
• Image of rms/mean for the 53 second observation (.gif)
• the rms/mean for each row (4096 spectra) was computed. The 320 of these (during the 53 sec file) where then made into and image.
• Since the rms/mean was done for each channel, there was no need for a bandpass  correction.
  
• The image was scaled to emphasize the rfi.
• 1232-1255 - punta salinas.
  
• You see many 12 second  radar rotations with blanking whenever they point at AO.
• 1500-1530 looks like an image of the 1232-1255. This occurs right before and  right after blanking (when there beam is close to pointing at AO).It's [probably happening in the analog if/lo.
  
• 1270,1290 - remy radar.
• every 12 seconds (starting at 7.5 seconds) you see the line fatten and a horizontal line of spikes occur across the entire band. This is when the radar is pointing at AO. 
• 1330,1350 - faa radar.
• Creating spike across the band every 12 seconds when the radar points at AO.
• 1/2 way between when the radar points at ao, there are a set of widely spaced spikes that last for about 2 seconds
• 1409,1428 at  12-13 seconds, 1530,1560, 1610,1630,....
• The are either the horn spillover on the primary when the radar points away from us, or it is a plane that is passing by. If it is a plane, then the position in  time will change. These look more like some intermods in the analog system.;
• Image of rms/mean for 53 second observation scaled to see weaker changes (.gif)
• This image is the same as the previous. It uses histogram equalization for the scaling.

Summary

• Need to check if some of the bandpass ripples are coming from cables.
• The noise strength is  following the radiometer equation (at least at the 40 usec level..).
• The FAA, and Remy radar are causing spikes throughout the spectrum when the radar  points at AO.
• The online monitoring was showing min,max spikes across the spectra. They are probably coming from these rows when the radar points at AO.
  
• This is either the A/D or (more probably) overflows  in the butterflys for the fft.
• changing the down shift rate in the butterflies (form 0xaaaaaaaa) to more shifts could help
• in the butterflies, noise grows as sqrt(N) where N is the number of butterflies done. For sine waves or radars, it grows as N.
• If  more downshifts are done, then the 8 bits that are extracted from the 18 bits of the fpga may need to be changed.
• This test used 2048 channels and incoherent spectra. For coherent dedispersion the fft length is shorter (512) so there are fewer butterflies used. This should help this problem.
  
• The rms noise was set to around 22 counts. You could lower  this a little without hitting the digitization  noise threshold.
• We need to measure the max input voltage at the A/D to see if the A/D is  getting zappped.
• The radars are also causing intermods that look like they are coming from the if/lo. Need to check if the two amps before puppi's a/d have enough dynamic range.
  
• Since there is only one analog mixing stage we can probably predict where these should show up in the 1-2 ghz IF  or from the digitizer.
  

home_~phil