p3058: puppi radar saturation
On 21march18 p3058 (using a p2030 time slot) did
a confirmation run using lbw and puppi.
They reported very strong radar interference and questioned
whether the radars had stopped blanking
in the AO direction.
Their setup was:
- lbandwide, cfr 1380 MHz, 1150 to 1750.
- 10 GHz upconverter mixing to 1st IF
- IF band 800 MHz -> 1600 MHz gets sent to puppi (only 1000
-> 1600 MHz IF freq has signal)
- The table below shows the layout of the first IF used by
puppi (this does not used the high freq puppi mixer).
|Edge of band sent to puppi. (out of 1-2
GHz IF band)
|Edge of 1-2 GHz IF band filter
|center of 800-1600 MHz IF band sent to
|Upper edge of LBAND wide sent to puppi
(rf filter defines this).
- puppi ran with 40.96 usec sampling, 2048 channels,
polarizations summed, search mode, with 4096 spectra/row.
I Looked at two files recorded during the run:
- puppi_58198_B1924+16_0014_0001.fits (5 Gbytes) (120
- puppi_58198_C1845+0016_1684_0017_0001.fits (11Gbytes)
Average Spectra for
each file (top)
The first plots show the average spectra
for each file (.ps) (.pdf)
- Each row was first averaged (to .168 seconds)
- The min,max, and average was then computed for each row of the
- black trace: average spectra
- blue trace : min hold spectra. (using the row averages)
- red trace: max hold spectra ( using the row averages).
- the max hold for the 40.96 usec spectra would have much
- The band falloff at 1150 MHz is from the 1-2GHz IF filter (see
- Top frame: puppi_58198_B1924+16_0014_0001.fits
- the 1175, 1127, and 1575 spikes are caused by a GPS
- GPS prn 24. it came within 7.7 degrees of the AO beam
during the first file.
- The large birdie at 1050 MHz is interesting since it is
outside the IF1 1-2 GHz filter. So it is created after the
- Bottom frame: puppi_58198_C1845+0016_1684_0017_0001.fits
- similar rfi, but the GPS signal is not as strong.
Dynamic spectra for
each file. (top)
The dynamic spectra were computed for each
- Each row was first averaged (giving .168 secs resolution).
- the median bandpass was divided into each spectra before
- The dynamic range of the image was set to 5 sigma (peak
display level at 5 sigma).
Dynamic spectra for B1924+16
Dynamic spectra for C1845+0016
- The FAA frequencies have been flagged in green
- the punta borinquen frequencies are flagged in red.
- A dashed horizontal green line is plotted (at 54 seconds) when
the FAA 12 sec rotation pointed at AO.
- A dashed horizontal red line is plotted (at 54.5 seconds) when
the punta borinquen 12 sec rotation pointed at AO.
- When the radar points at AO, there are a large number birdies
(harmonics or intermods) that are being created.
- Similar rfi present in this file.
- You can also see GPS L3 (1381MHz) in this image.
radars blank when they point in the AO direction. (top)
There is an agreement with the FAA, DOD, and
homeland security to blank the FAA and Punta Borinquen
radars when they point at Arecibo. The memo says they will blank
for 4.25 degrees (or .142 seconds during their 12 second
rotation). Without this blanking, the AO receivers would go into
The next plot shows a times series of the radar
frequencies as they point at AO (.ps) (.pdf)
- 1.5 seconds of data was taken from the C1845+0016 file.
- The total power was computed for 2MHz about each radar
frequency (they are chirped by 2MHz).
- This was done at 40.96 usec sampling resolution.
- This time period corresponds to a blanking period for the FAA
and the PuntaBorinquen radar.
- The bottom 4 traces are from the FAA radar (offsets are added
for display purposes).
- The top 4 traces are from the punta borinquen radar.
- The blanking period for each lasts for 1.4 seconds
(or 4.2 degrees).
- So they are blanking the radars with the agreed upon
- If the blanking were not present, the strength of the radar
would increase many times (expect for puppi saturating,,,)
- The spiky lines after the blanking is from the 2-3 millisecond
ipps of the radar.
- The plots also show that the blanking is aligned with the
radar beam pointing at AO.
- If the alignment was off, then the pulses would be stronger
on 1 side of the blanking.
- The plots also show that there are no extremely strong spikes
outside the blanking area.
Dynamic spectra were made (at full time resolution) for an FAA
and then PuntaBoriquen blanking period.
Dynamic spectra during FAA
Dynamic spectra during Punta
Borinquen blanking (.png)
- The images are made using the 40.96 usec resolution.. but for
display, the time axis was smooth/decimated by 8 giving 320 usec
- the millisecond spacing between the horizontal lines are the
ipps from the radar.
- The FAA frequencies have been flagged in green
- The harmonics/intermods are spaced by about 50MHz. This is
probably related to where the rfi falls in the IF band.
- The harmonics/intermods get a little weaker as you move away
from the blanking.. since the beam is not pointing directly at
- The punta Borinquen frequencies have been flagged in red.
- looking at the images, the problem is definitely coming from
the FAA and PuntaBorinquen radars.
Looking at 1
Radar pulse from the FAA radar: (top)
A summary of the radar properties for the FAA
- 4 frequencies. Each frequency chirped by 2MHz.
- 3 different ipps: [2746.9166,3154.7360,3473.4905] usecs
- each ipp repeated 5 times, then it moves to the next ipp. so
15 ipps before repeating.
- after 30 rotation periods 5 ipps are dropped.
- 12 second rotation period.
- The frequencies, transmission order and duration for each
frequency is shown below.
in radar pulse
|cumulative time at start
The final plot shows
data taken during a single radar pulse (.ps) (.pdf)
- page 1: total power and spectra during the pulse.
- Top Frame: total power time series for each FAA freq during
a radar pulse
- 1500 usecs are shown (40.96 usec resolution).
- each color is a different radar freq.
- bottom to top is the transmission order.
- The black,red,green, blue vertical dashed lines are the
start of each pulse (using the arrival of the first pulse as
the start time).
- The purple dashed lines show the time period used for the
spectra in the bottom frame.
- You can see there is some overlap in the pulses.
- part of this is from the 40usec sampling
- But some overlap is expected because multi path
scattering will delay the pulse
- So there is a possibility of intermods occurring between
different frequencies of the radar
- this would not be possible if there was no overlap in
- Bottom frame: the spectra within the purple lines of frame 1
- The number of birdies increases during the overlap of the
first and second frequencies.
- Page 2: single spectra during middle of FAA radar pulse
- this is the start of the 1349.59 Frequency.
- the birdies are evenly spaced across the band.
- The birdie strength is the same at 1050 MHz as 1350 MHz.
- The 1-2 GHz IF filter starts to fall off at 1200 MHz RF
frequency. By 1050 MHz RF frequency, it is down by
10db or more
- This shows that the intermods/harmonics are being created
after the 1-2 GHz IF filter.
- my guess is that they are being created by the puppi A/D
(8bits) or maybe the pfb (it's easier for a strong rfi to
overflow in the butterflies if you only downshift the fpga
registers every other butterfly.
- lband wide data taken on 21mar2018 using puppi showed strong
harmonics/intermods being created in the output spectra.
- The FAA and PuntaSalinas radar signals are causing this
- The data show that the 2 radars are blanking for 4.2
degrees of their rotation cycle.
- The blanking is also aligned with the azimuth rotation of the
- multi path scattering causes the frequencies from a single
radar to overlap in time.. causing extra harmonics/intermods.
- The intermods/harmonics are evenly spaced across the 800 MHz
band of puppi.
- Even in the regions that are cutoff by the 1-2GHz IF filter.
- this shows that the saturation is occurring after the IF
- Either in the puppi A/D or maybe the PFB.
- The cima log file for the experiment had:
- change_puppi_scales "set" "04000000"
- I think this modifies the shift rate in the PFB butterflies
(although I'm not sure..)
- I wonder if this setting causes more or less attenuation to
be inserted (when the adjust power is run).