• dynamic spectra
  
• average on,off spectra
• the comb period.
  

• dynamic spectra
• average on,off spectra
  

Summary

Introduction/timeline (top)

• 19feb17 p1693 data shows no 40khz comb
• 28feb17 p1693 data shows strong 40 Khz comb
  
• Note.. these were inspected after the fact..
  
• 15may17 az swings showed 40khz comb in 327 receiver.
  
• 01jun17 : rfi tests done turning sband console on,off while taking data showed sband console causing the comb.
• victor worked to get rid of the emissions from the cabinet.
  
• 05jun17   p3166 .. scott ransom reports rfi problems with 327  puppi data
• 29may17 intermittent strong rfi  in data
• 04jun17   strong rfi not present.
• By the time i got around to looking at his data, it was no longer available at AO.
  

01jun17 turning the sband console on,off

• az,za fixed.
• 327 receiver, cfr 327, 750 narrow band mixer
• mock spectrometer, 7 x 8 Mhz bands, 8192 channels/band (976 Hz channel resolution).
• while taking data turn the sband cabinet power off and then back on (about 100 seconds off)

Before changes were made to the cabinet:

• 490 1 second averaged spectra were taken
• we started with the cabinet closed, and on
• at 315 seconds, the power to the sband cabinet was turned off
• at 420 seconds, the power to the sband cabinet was turned back on

The dynamic spectra:

    The dynamic spectra were made for each 8mhz band:

• The median bandpass  when cabinet was off was computed.
  
• 20the order harmonic was fit to this median bandpass
• flatten each of the spectra with the fit.
  
• remove tsys from each spectra using the median
• average the two pols.
• I placed horizontal lines where the cabinet was powered on ,off
• powering the cabinet on,off would take a few second for the times are not exact.

• The comb goes away during 320 to 420 seconds in the images. This is when the cabinet is turned off.
• It comes back when the cabinet is turned back on.
• Note:
• strong rfi with a 200 khz spacing occurred until 140 seconds.
• coincidentally , victor and his crew opened the door to the transmitter room (in the dome) around 133 seconds.. This may just be a coincidence. Not sure how long they left the door open (i couldn't see the door on the camera).
• You'd expect things to get worse, not better when the door was openned.. (i wonder if the turn on or off the fan when they entered?

Average on,off spectra before fixes:

  For each 8 MHz  band, i averaged when the power was all the way on, and when it was all the way off.. i tried to leave out the transition periods.

The plots show the average spectra for cabinet on and off (.ps) (.pdf ):

•  Each page shows 1 8MHz band.

• Top frame: vertical scale is 0 to 10 times  Tsys
• Bottom frame blows up the vertical scale to 0 to .1Tsys
• The on time avg is 380   seconds. the off time avg is 105 seconds.
  

• Black is with the cabinet off, red is with the cabinet powered on.
• I plotted the red (on) and then over plotted the black (off) so the on could not hide the off.
  
• If you see a red line, with a black line over it, then that rfi did not come from the sband console
• Red lines sticking up by themselves are coming from the sband console.
• The on comb gets up to about 10 times Tsys (around 318 Mhz)
• the blowup to .1 tsys shows that comb covering the entire 327 band.

Comb period:

• abs(fft(spc)) is plotted.
  
• the vertical scale is the strength (power)
• the horizontal scale is set to the freq (1/period) of the combs.
• the plot for each 8 MHz band is offset for display.
  
• The comb is at 39.7 KHz.
  

After repairs made to the cabinet:

    Victor and his crew worked on repairing the sband console shielding on 170601. After the repairs, we repeated the sband on/off console tests using the 327 mhz rcvr and the mock spectrometer (see above for setup).

Dynamic spectra while cabinet turned on/off.

• these are the dynamic spectra made after the cabinet was repaired.
• I placed horizontal lines where the cabinet was powered on ,off
• powering the cabinet on,off would take a few seconds.
• i flagged where the cabinet was completely off..

• No obvious comb remains.
• there is a chirped signal spaced by about 1.25 or 2.5 Mhz that is seen throughout the band. It remains on the entire time, so it is not coming from the sband console.
  

Comparing cabinet on with cabinet off after repairs:

    For each band i averaged when the power was all the way on, and when it was all the way off.. i tried to leave out the transition periods.

The plots show the rfi when the cabinet was on and off (.ps) (.pdf)

• Each page shows 1 8MHz band.
• Top frame: vertical scale is 0 to 10 Tsys
• Bottom frame blows up the vertical scale to 0 to .2 Tsys
• The on time avg is 246 seconds. the off time avg is 122 seconds.
  

• Black is with the cabinet off, red is with the cabinet powered on.
• I plotted the red (on) and then over plotted the black (off).
• If you see a red line, with a black line over it, then that rfi did not come from the sband console
• Red lines sticking up by themselves are coming from the sband console.
• The vertical axis is in units of Tsys.. the plots are clipped to 1.5 Tsys
• Nothing obvious sticks up.
• there are a few red on peaks with no black off peaks.
  
• Looking at the dynamic spectra, these are rfi bursts that happened in just one of the on sets. They don't appear in both ons. To get around this we could have done more on,off cycles...
  

Summary: (top)

• A 40 Khz comb in the 327 receiver was coming from the sband console.
• The  comb got up to 10 * tsys.
• at the .1*tsys level, it was continuous across the 50 Mhz of the receiver.
• I don't know if scott's problem was from this comb (i didn't have his data to look at..)
• i wouldn't expect the comb to be time variable.. he said the problem occurred in the middle of his observation.
• His observation occurred during the day on memorial day.. The visitor center had lots of visitors.. could have been some device on the observation deck..
  
• Looking back at p1693 pulsar search data (using the 327 receiver)
• The comb was not present on 18feb17
• the comb was present on 28feb17.
• So we went about 4.5 months with this comb in the receiver...
• The comb doesn't show up in the fractional occurrence of rfi on telescope data .. because the rfi is not time variable..
• pulsar people seem not to care (or at least observer)  this comb.. i guess they never look at the spectra before dedispersing..
  
• spectral line observers have given up on the 327 receiver (because of the rfi...a lot of it is internal..).
  
• After the repairs, the comb is no longer visible in a 100 sec integration with 1 khz resolution.
• The sband console has caused the same problems in:
• jan 2016
• jan 2009
• apr 2006
• dec 2003
• We should change the cabinet to one that is more rfi tight.
• Whenever the sband crew opens/ works in the cabinet, we need to do an rfi check to make sure that things were put back with no leakage.
  

processing: x101/170601/ rfi327_before.pro rfi327_after.pro

home_~phil