Intro

    On 11mar22 the new sband filter was installed in the postamp chassis in the pedestal. The sband inputs to the postamp chassis were connected (with no pads at the input). This allowed us to take data with the sband system without saturating the sband receiver chain.
    On 22mar22 3 azimuth swings were done with the sband receiver to look at the azimuth dependence of the rfi.

---

Setup:

• The receiver was set to a center frequency of 2270 MHz
• if/lo attenuators were set to:
• udc: 10,10 db (pola,b)
• if2  :  -8,-12  db (8,12db of gain)
• A single mock spectrometer was used to record the data
• 160MHz bw polA,B recorded
  
• .1 second integration
• 8192 freq channels.
• The power levels were adjusted to a rms of 30 count=1 sigma
  
• turns out we ended up with about 36 counts since we were pointed at -90 degrees which had some rfi.
• The telescope pattern was:
• The elevation was kept constant at 30 deg
• the azimuth rate was set to 1deg/second.
• 3 azimuth swings were done:
• swing 1: -100 to 280 deg az CW
• swing 2:  280 to -100 deg az CCW
• swing 3:  -100 to 280 deg az CW swing
• datataking started at 13:47
  
• each swing took 380 seconds
• we finished at 14:07

---

Processing the data"

• Each azswing was read in with  the idl routine masazswinginp().
• input the file
• select the data between -90 and 270 degrees azimuth
• interpolate the azimuth position to the center of each .1 second spectra
• the azimuth is recorded once a second.
• add the polA,B  spectra together.
  
• return the spectra in increasing azimuth (so the 3 swings can be easily  over plotted).
• generate a dynamic spectra for each azswing.
• each data set had 8192 channels by 3000 spectra.
• these were averaged to generate an image that could be displayed
• average by 8 in freq and 5 in time ending up with a  1024 channels by 750 spectra
  
• 390 Khz freq resolution, .5 sec time resolution (or .5 deg  in az since we moved at 1 deg/sec)
• The spectra were bandpass corrected by dividing by the median spectra.
• the median power between 2268 and 2315MHz (little rfi)  was computed for each spectra and then each spectra was divided by this value
  
• (to remove gain variations and continuum sources)
• Each dynamic spectra was displayed with low and high sensitivity (to see the location of the strong, weak rfi.
• The rms/mean by frequency channel was computed for each az swing (3000 spectra).
• larger rms/mean values show channels with rfi.
• The frequency channel with rfi were located and the then total power was computed within this freq range for each spectra.
• The rfi total power was then plotted vs azimuth.
• If the rfi repeated at the same azimuth for the 3 az swings, then this is probably the direction for the rfi.
• A final test was to check to see if we had any compression in the measured signals.
• the total power in a spectral region with little rfi was computed and then plotted vs azimuth.
• if we saw a dip in this power then we probably had compression of the signal.

---

Looking at the data:

 Dynamic spectra

  The table below contains the dynamic spectra for the 3 azimuth. For each swing a high and low sensitivity image is made.

azSwing	HighSensitivty	low sensitivity
1 (CW)	.png	.png
2 (CCW)	.png	.png
3 (CW)	.png	.png

• The rf filter at sband  passes 2205 to 2345 (more info)
• The strongest rfi is 2320 -> 2355 (around az=255 and an azimuth to 195)
  
•  rfi below 2200 is also strong
• you can see a few horizontal black stripes. This is probably compression/saturation.
• On 11mar22 spectra were measured at the input to the postamp chassis (before the rf  filter)
• This shows the rfi before the filter. The rfi 2100 to 2200 mhz is huge (more info)
  

rms/mean and peak hold spectra, saturation.

  
 the plot shows the rms/mean by channel for the spectra taken in the first azimuth swing (.ps) (.pdf)

• There are two pages  of  plots.
• page 1 includes all spectra in the azimuth swing
• page 2 excludes spectra where a dark horizontal line was seen it the dynamic spectra.This is normally caused by compression/saturation.Plot
  
•  Top: rms/mean by freq channel
• the rms/mean by freq channel was computed using the 3000 spectra of the first azimuth swing.
• Black is polA, red is polB, green is the expected rms
• 19Khz channel width,.1 second integration.
• the non rfi locations are close to the expected rms/mean.
• the rfi at 2230 has flattened out the rms/mean for pol.
• Page 1:
  
• The black spikes (a few mhz wide) are not real. They are being created by the saturation at 2230 MHz.
• Page 2: these black spikes are not seen when saturated spectra are excluded.
  
• Bottom: The peak hold spectra for the first azimuth swing.
• The peak value in each freq channel for the first azimuth swing.
• I've divided by the median value so the vertical units should be close to Tsys.
• You can see that the 2230 rfi in polA has saturated.
• Page 2 does a better job of showing the rfi (outside the 255 MHz region).
  

    The 2nd plot looks for saturation of the signal (.ps) (.pdf)

• Top: total power 2272-2280 MHz vs az (using azswing 1)
• this region had little rfi
• Around an azimuth of 255 the level drops by about 4% probably due to saturation somewhere else in the band
• Bottom: 10  .1 second spectra around az=255 deg
• the signal is saturating at 2330 MHz.
• The vertical scale is mock 32 bit accumulator counts.
• It has hit the limit of 2e9 (largest positive number in 32 bit signed value)
• Looking at the status bits from the mock spectrometer we also see the  polyPhaseFilterbank overflowing with az 250-255.
• the rfi we see spread across dynamic spectra at az=255 is  not real. It is being caused by the saturation/overflow.
• the black 5mhz spike across the peak hold are also being created by this saturation.
  
• We don't  know if this rfi is saturating. the analog part.
• The tsys baseline is 1e8
• the rfi is at least 2e9 in 5MHz. the entire band is 130 MHz
• so 5/130 *20 = .8  so this rfi is at least equaling the normal power out of the filter.

plotting individual rfi birdies vs azimuth 

    The plot shows the azimuth dependence of various birdies (.ps)  (.pdf)

• 21 different rfi birdies were selected from the azswing1 dynamic spectra
•  the total power was computed, and then plotted vs azimuth.
• the value from the 3 azimuth swings are over plotted
• black: swing 1, red: swing2, green: swing3
• If the rfi  is at a fixed location then is will probably show up at the same azimuth for each az swing
• assuming it is not too time variable.
• the table below summarizes  the results
• Some birdie constant birdies were not include (seen in the peak hold spectra)
• 2270 -this is the dc component from the mock a/d converters
• 2250 - this is not seen very clearly in the dynamic spectra since we bandpass correct with the median spectra.
  
• 
  

• freq	az	repeats in az	PkVal tsys	Notes
  2192.9-2199	120	yes	11	strongest at az=120, but  is large for the entire swnig the 3 azswings show the same azimuth dependence.
  2202.92-2205.31 2207.19-2211.87 2250.77-2254.05	255	not real		these are intermods created in our system by saturation of the 2330 rfi
  2217.8-2218.9	103	no		very narrow in freq
  2225.82-2227. 2227-2229	-40	maybe all az	2	strong in first az swing but seen in the other dynamic spectra, but much weaker
  2230.6-2233.6	95	yes	1.2	also seen at other az.
  2338.102239.7	-40	yes	1.1	1st swing is a little lower in az.
  2240.1-2241.24 2242.18-2242.8 2243.6-2244.9	102 120	yes	2max	these are 3 narrow spikes in freq that occur together.
  2267.2-2268.9	-	no	.5	only occurs in azswing 1 at various azimuths.
  2305-2308.6	-	not real	1	look like they are coming from rfi > 2320
  2312.2-2313.4	-	-	-	narrow tone. looks like always present.
  2319.2-2324.51 2327-2332 2332.6-2335.9 2336.2-2341.2 2341.3-2344.8	255 195 5	yes	>1000	sirius sat1 (4mhz) sirius sat2 (4mhz) xm sat1a,sat2a  (1.84MHz each) xm terA (2.53MHz) may just be leakage from adjbands xm  sat2b,sat2b  (1.84 Mhzeach) 2327 - 2332 is the strongest > 1000 Tsys. It is causing intermods in the mock spectrometer. The other signals are around 2 and 20. sirius-xm freq allocations (more info)

---

Summary

• 3 azimuth swings were done from -90 to +270 azimuth at 1 deg/sec
• strong rfi was seen above 2320 MHz around an azimuth of 195 and 255.
  
• they repeated for all 3 azimuths
• 2320-2324 MHz 5*tsys
• 2327-2332 MHz  > 1000*Tsys. saturated the mock spectrometer
• 2332-2236 MHz about 20*Tsys
• 2336-2341 MHz about 2*Tsys
• 2341-2343 MHz about 200*Tsys
• 2320-2345 is allocated to siriusxm digital radio.
• The sirius satellites are at 85 and 115 degrees west. these may map to the 12m azimuth's of 195 and 255.
• If we track the satellites, we may get a larger signal.
• the 2327-2332 rfi (sirius band) caused the mock spectrometer to saturate the 32 bit power accumulator
• we also so overflows in the butterfly stages of the pfb
• we don't know if the analog system went into compression.
  
• Todo: get a tle for the sirius xm satellites and then see how close we have to be to the satellites to saturate. the system.
  

processing: x101/220322/sbazswing.pro