Links:
Setup
220608 map
dynamic spectra
jun-jul7 occurrence  of saturation
summary

Intro

    A group (sscontin) is doing xband maps with the 12meter in the Cygnus region.  The maps are using the hardware winking cal. In jun22 they reported a problem with the map data. They saw  dips in the total power along strips of the map. and their guess was that there was an issue in the software extraction of  the calon,offs.

   I looked at the maps taken on 8jun22 with the reported data. I then examined all 2x2 deg maps taken jun-jul22

---

Setup:

    The mapping project has the following setup:

• The xb receiver was used.
• 7 172MHz bands  cover  1GH at xband
• 2x2 degree  maps are done (i ignored the 1x1 degree maps)
• 2 types of map are done to allow for basketweaving:
  
• drive in ra, step in dec
• drive in dec, stop in ra
• The steps are always 1/2 beam=5Amin (great circle).
• The driven direction covers 2 deg in 60 seconds.
• The hardware winking cal  is used during each strip
• The cal has a 25Hz cycle: 20ms calon, followed by 20ms cal off
• It is synchronized to the hardware 1 second tick.
• Spectra are sampled at 2ms/spectra
• 10 spectra calon, followed by 10 spectra cal off
• The calOn, calOff spectra are separated in software.
• 8 samples, 8 samples off are averaged on each 40ms calcyle
• This excludes the 2 ms spectra before and after each cal transition.

---

Looking at the 08jun22 map

    The first plots show the data from the 8jun22 map (.ps) (.pdf)

• page 1: total power vs time for each strip (8219MHz band)
• each color is a different strip in the map (25 strips total)
• The vertical scale is degK.
  
• A baseline has been removed from each strip.
• an offset has been added to each strip for display
• Top: polA
• bottom:polB
• You can see the negative going spikes that are in each strip of the map.
  
• Page 2: totpwr vs time for strip0 for all  7 freq bands
• Each color is the total power from a different frequency band
• an offset has been added for display.
• Top: PolA
• bottom: PolB
• You can see that the spikes occur every four seconds.
• the cal cycle is 40 milliseconds.
  
• It if very unlikely that the spikes are a calon,off problem.
• Looking at the 5th and 6th bands (8789,8991 MHz)
  
• the negative going spikes are now large positive going spikes.the
• The negative spikes are caused by saturation (not cal transition problems).
• Since all 7 bands show a problem:
• The problem is occurring before the 1 GHz is split into 7 bands
  
• the IF2 and the mocks are probably not saturating.
• The problem is most likely in the udc up/down converter (more info)
• The udc hits 1db compression at -35 dbm input.
• Page 3: spectra strip0 .2 to .3 seconds
• this over plots the spectra for all 7 bands during the first dropout
• The rfi goes from 8730 to 9000 MHz.
• Page 4 : totpwr and spectra blowup
• the total power and spectra have a 2 millisecond time resolution.
  
• Top: total power blowup for 8700 to 8900 MHz]
• black is polA, red is polB
• the rfi last for 39 milliseconds.
• the small square wave is the cal going on,off every 20 milliseconds.
• Bottom: spectra during the rfi pulse.
• Each color is a different 2 millisecond spectra
• An offset has been added for display
• polA and polB have been averaged.

Dynamic spectra for strip 0 of the 8jun22 map

    The image shows the dynamic spectra for strip0 of the 8jun22 map (.png)

• The x axis is frequency, the y axis is time (in seconds)
  
• the 60sec/.002=30000 spectra of strip0 were used to make the dynamic spectra.
• the cal off portion of the data was used.
  
• the 7 freq bands of each sample were interpolated to a  single spectra.
• freq 8710 to 9010 where then selected for the image.
• To fit on the display window the  time axis was smoothed/decimated by 40.
• The 4 second time periodicity is easily seen.
• The frequency covers 8732 to 8968 MHz
• It like there are 12 20MHz channels with a guard band around each one.
  

---

How often does the rfi occur?

    All 2x2 maps taken 08jul22 thru 16jul22 were examined (39 2x2 maps) to see how often the rfi occurs.

    The plots show how often the rfi occurred (.ps) (.pdf):

• Page 1: Strip 0 from each of the 39 maps was examined using the 8219 MHz band.
  
• the 20ms  cal off was averaged and used for the plots (1500  20ms avgs in 60 secs).
• each strip was flattened by a median filtered version of each strip (filter length=7 samples).
• Top: Total power vs time from strip 0 of the 39 maps (polA)
• bottom: total power vs time for strip 0 of the 39 maps (polB)
• The vertical scale is Tsys units.
• an offset was added to each strip for display
• The negative going spikes show how often the compression occurred.
• Page 2: the 9 maps with saturation are plotted.
• top PolA
• bottom: polB
• 9 out of 39 maps had saturation.
• Page 3: when and where the saturation occurred.
• Top: Az ,Elevation for 1st strip of each map
• The black * have no saturation
• the red * have saturation.
• there does not appear to be an az,el dependence of the saturation
  
• Bottom: when did the saturation occur?
• the minimum value for each median filtered strip was plotted vs date.
• the black * had no saturation.
• the red * had saturation

---

Summary:

• The Cygnus mapping at xband is seeing saturation in the 1 GHz band.
• The saturation is occurring  in all 7 bands, so it is before the mocks and the 2nd IF
• The causes dips of up to 4% of Tsys.
• It occurred in 9 out of the 39 maps examined.
• No az,el dependence was seen of the region covered.
• The saturation is being caused by rfi in 8732 to 8968 MHz
• It has a 4 second periodicity
• lasts for about 39 milliseconds (at least its strongest part)
  
• has 12 20Mhzz channels covering the freq range with guard bands between the channels.
• The FCC allocations for this freq range are:
• primary:  8750-8850:airborne services:
  
• 47.87 airborne  doppler radar.
  
• 47.87. freq table 47.87.173
• secondary: 8500-9000 private land mobile radio services , radio location
  
• 47.90. freq table
• 49.90.103 share but secondary to airborne doppler
• The frequency range of the rfi extends outside of the airborne services.
• The rfi is probably private land mobile
  
• the 20MHz channels look a lot like other communication/internet provider bands with the 20MHz bw and guard channels
• What can we do about it?
• If the saturation is in the UDC  then we can increase the attenuator between the FO rcvr and the UDC input.
• We need to verify that was remain well above the noise floor of the UDC
• We could monitor the power level at the FO rcvr output to see how high it gets (if we can catch the rfi when it is occurring frequently)
  

---

processing: x101/220608/mapsygsaturation.pro,cygmapst.pro