Alfa (p2030) 12 second compression is faa radar.

     The alfa pulsar search (p2030) saw strong 12 second periodicity's on 01aug05.  They sample data at 64 usec using 100 Mhz (1370 to 1470)  with 256 channels. They add the polarization's before storing the data. I looked at one of the 2 gb files (about 130 seconds of data) where the problem was the worse: p2030.G35.21+01.00.N.wapp1.53584.0004. This contains pixel 0 and pixel 1 (i ignored pixel 1 since pixel 1b is dead). The plots show the total power versus time (.pdf):

• Top: The total power (100 Mhz) versus time. The data has been smoothed to 64 milliseconds. The y axis is in units of Tsys. There is a large spike every 12 seconds with a long recovery time (up to 12 seconds). The dashed red lines are the region blown up in the 2nd plot
• 2ndPlot: A blowup around 30 to 31 seconds. This has the full 64 usecond resolution. The  negative going pulses are where the system is compressed. They start to get large and then at 30.34 seconds the baseline increases by 40%. This is probably where the radar beam sweeps in front of the observatory (this normally takes about 80 milliseconds). The baseline does not come back to the original value for up to 12 seconds. The dashed lines show the portion plotted in fig 3.
• 3rdPlot: This has .2 seconds of data. You can see the individual ipps of the radar where they saturate the system.
• 4thPlot: I computed the difference between adjacent negative going spikes in figure 3 and then plotted vs time. Different radars  have different sequences of ipps. The faa airport radar at 1350/1330 Mhz has 5 different ipps (see rdrReport(.ps)  pg 4). The colored lines are the 5 ipps for the faa radar. They agree with the differences between the negative going spikes. This shows that this compression is being caused by the faa airport radar. The radar frequencies are 1330 and 1350 Mhz so both of the frequencies are out of band for the digitizer.

• TopPlot: The average spectra for the 130 seconds. The central peak is mainly galactic hydrogen. I split the spectra into 8 12.5 Mhz junks and computed the total power for each of these versus time (with 64 Usec resolution). The colored dashed lines show the center of each 12.5 Mhz band.
• 2ndPlot: The total power vs time for the 8 12.5 Mhz band in  the 100 Mhz spectra. This is at the 30 second time when the baseline changed. I normalized each 12.5Mhz total power strip to the first 20 Milliseconds (before the baseline started to change). You can see that all 8 of the 12.5 Mhz sections do exactly the same thing so the change is broad band.
• 3rdPlot: There were some spikes at the center of the band every 12 seconds. I plotted the power in the 1420 Mhz channel (black) when the baseline changed. The red line is the 100 Mhz total power (all 256 channels). When the baseline started to change, the central channel increases (the adjacent channels show no large increase like this). This spike is a mixing product:
• lo:1670, rfi:1330.... 3*rfi - 2*lo = (650Mhz).
The 750 Mhz lo pass filter after the mixer lets this intermod pass into the 1st IF. The normal band is centered at 250 Mhz. It is digitally down converted to 0-100 Mhz by the digitizer. The birdie at 650 Mhz will also get digitally down converted to the center of the 100 Mhz band by the digitizers.

home_~phil