Setup

• 327 receiver
• puppi with 68.75 MHz bw (2816 channels), polarizations added,  81.925 usecs  integration per spectra, incoherent search mode, 8 bit data recorded.
• 15 files were taken with puppi scan numbers: 0174 to 0189 (scan 184 is missing).
  
• file 1, scan 174  had 60 seconds of data (on pulsar 1713+0747).
  
• files 2-15, scans 175-189 had 15 minutes of data (searching).
  
• For these files i used the last 46.29 seconds of data (the 2nd 1.5 gbyte file of the 15 minutes)

Average and rms/mean of spectra

    The average spectra and rms/mean was computed for each row of the first fits file (scan 174)

• I think that the puppi input data was adjusted to have a sigma of 20 counts.
  
• Each row in the fits file contained 8192  81 usec spectra (.671 seconds of data total).
• There were 90 rows in the file.
  
• The average spectra were computed by averaging the 8192 spectra of each row (giving 90 .671 sec averaged spectra)
• The rms/mean by channel was computed for each row
  
• The rms was computed for the 8192 samples for each channel and then divided by the mean value.
• The rms/mean should match the radiometer equation: deltaT/T=1/sqrt(bw*time*2)
  
• The factor of 2 comes from polarization adding.
  

• Top : average spectra for each row (.671 seconds) are over plotted.
• The vertical scale is 8 bit counts of the output spectra. The average value of the spectra is around 10-15 counts
• The filter shape comes from the 52 MHz rf filter and the from the 60 MHz IF filter (which has steeper edges).
  
• Bottom: rms/mean for each channel.
• The rms was computed along each channel of the 8192 spectra of each row.
  
• This data has 81.925 usecs time resolution and .0.024 frequency resolution (assuming rectangular filters)
  
• The rms is divided by the mean value of each channel. The value of this plot should correspond to the radiometer equation:
• deltaT/T= 1./sqrt(bw*time)
• bw=0.024414062E6 hz
• time= 81.925E-6 secs * 2
• (times 2 since we've added the polarizations
• The red dashed line shows the expected rms/mean the bandwidth and integration time of the data.
• The measured value may be a bit higher that the computed values because the polyphase fitlerbank may make the filters a bit narrower.
  
• The spectralMean* (rms/mean) give the number of  counts of noise in each channel.
• For the center of the noise on top of the spectra has about 8 counts (15*.53).
• Towards the edge of the band the noise has about (2*.6)=1.2 counts. This will start to add some noise to any processing

Dynamic spectra bandpass correction:median vs Fit.

• Median bandpass corrected Dynamic spectra of file 1 (.gif)
• Using a median bandpass correction will do a good job of flattening the image.
• Any constant rfi will also be removed.
• Fitting a bandpass correction to the Dynamic spectra for file 1 (.gif)
• a 23 order harmonic fit was done to the median bandpass and then used for the bandpass correction.
  
• Any harmonic structure greater than 23 cycles will remain in the image
• Constant rfi will not be removed from the dynamic spectra.
  

File 1: rfi with 10 sec period.

• Dynamic spectra were made with the 81 usec spectra.
  
• Blowup of 1st pulse in the file (.gif)
• The pulse lasts for about 160 milliseconds
• The band pass correction by the mean value caused the dark stripe
• At the start of the pulse, the frequency sweeps from above 360 MHz down to 353.4 where it then stabilizes.
  
• pulses 1-3 of the file (.gif)
• pulses 1-3 show the sweeping at the start of each pulse
  
• pulses 4-6 of the file (.gif)
• pulses 4-6 continue with the sweeping at the start of each pulse.
• The plots shows 100 spectra (81.925 usecs each) at the start of the first pulse (.pdf)
• Top frame. the value gets close to the maximum value of the 8 bit data.
• Bottom frame: the spectra have been offset for display
• You can see the signal sweeping in from 360 MHz
• The slew rate is -.001888 MHz/Usecond

Dynamic spectra of files 1-15

• dynamic spectra for files 1-5 (.gif)
• File 2 shows the 10 sec period rfi also appearing at 333.5 MHz.
  
• dynamic spectra for files 6-10 (.gif)
• 3rd file from the bottom:
•  353.4 pulse spreads out in freq. It is probably saturating
• 333.5 also appears at 10 sec period
  
• dynamic spectra for files 11-15 (.gif)
• 2nd file from the bottom:
• 297.36, 353.10 MHz show the same 10 second periodicity
• 353.88 has a 10 second periodicity that appears a little before the 353.10 pulse.

Summary

• Mean values:
• center of the band has about 15 counts (out of 255).
• edge of the band gets down to 1 or 2 counts
• Counts on the noise:
• Center of band about 8 counts
• edge of the band gets down to 1 or two counts
• The rms/mean for each channel is close to the expected value (from the radiometer equation).
  
• The bandpass shape is probably being determined by the 60 MHz narrow band IF filter.
• Bandpass correction:
• The median bandpass does a good job  of flattening dynamic spectra, but constant rfi gets removed
• Fitting a harmonic function to the median spectra and using that for the bandpass correction will not remove constant rfi (but it doesn't do as good a job flattening the image).
  
• RFI:

• 	Freq MHz	Notes
  rfi with 10 second period	353.16-353.55	seen in all 15 files pulse length about 160 millisecs sweeps from 360 to 353 MHz at start of pulse sweep rate -.00189 MHz/usec Value is probably overflowing the 8bit number
  	333.5	shows up with 353 in file 2,8
  	297.36,353.1	show up together with 10 sec period in file 12
  	353.88	file 12 10 sec period. Pulse arrives a little before 353.1
  fm station harmonics	308.7	3rd harmonic of 102.9
  	312.3	3rd harmonic of 104.1
  	1.2MHz comb	this is the difference between 102.9-104.1 probably intermods. The towers are about 100 meters apart (more info)
  AcUnits building 1	320	ac unit atop building 1. Probably a lot more
  AC unit on top of building2	lots of birdies	more info

• It will probably be hard to excise the 10 sec rfi completely because of the wide sweep at the beginning of each pulse. This looks like an oscillator or phase locked loop that is not doing very well.
  

 home_~phil