Alfa filter bank amps cause aerostat radar to compress
system.
may05
What was added to alfa and when:
A filter bank before the first mixers
was added to the alfa receiver for OH measurements. It has the standard
full bandwidth option and a set of filters for OH observing. A 10 db (actually
a mix of 10 and 14 db ) amplifiers were added before the filter bank to
improve the system temperature. These amplifiers were in the general rf
path and affected regular and OH observing. The relevant dates are:
-
mar05: filter bank installed but left in wide band
mode.
-
05apr05: 10db amps placed in pixel 2a,2b path in front of the filter
bank.
-
08apr05: 10db amps placed in front of filter bank for all alfa pixels.
-
12apr05: 6db attenuators placed before 10db amps.
-
25apr05: 10db amps are now only in the OH filter bath. Normal wide band
observations do not have any amps.
The 10db amplifiers improved the system performance
of the OH observations but caused the regular wide band observations to
compress when the radar signals were strong.
Looking at wapp data feb05 thru apr05 to look for compression:
The wapp pulsar data was recently placed in an idl data
archive. This allowed the search of the data taken by P2030 (100Mhz bandwidth,
cfr 1420Mhz, 64usec samples) between 01feb05 and 01may05 for signals of
the aerostat radar. The aerostat frequencies (1241 thru 1261 more
info) are outside the 1370-1470 Mhz band of the experiment but the
radar can compress the entire RF/IF chain. This can then be seen in the
p2030 data.
The processing steps were:
-
Make a total power time series model of the aerostat radar for a complete
12 second rotation (more
info).
-
Transform the model and keep the first 1 Khz of data (1/12 hz resolution).
-
From the model find all of the periods whose amplitude is at least
10% of the maximum amplitude. Normalize these periods to the maximum period
found. Use these periods to check the data taken by p2030.
-
For every file (1087 files) taken by P2030 between 01feb05 and 01may05
do the following:
-
create a total power time series for the first 12 seconds of the file (187500
points) for all 7 pixels.
-
transform each time series and then truncate to 1 Khz (1/12 hz resolution).
-
Remove the median value computed of this spectra (over 100 to 1000 Hz)
and then remove it.
-
multiply the model period amplitudes (a few hundred) by the corresponding
measured amplitudes and then compute the mean value.
-
The mean value will be large for data sets that are corrupted with aerostat
compression.
The following plots show the strength of the compression plotted
versus various parameters. When plotted versus date, this is the julian
date. The tick marks are centered on 12 noon utc for that day (8am AST).
The data is plotted versus sample or file found. The top plot is pixels
0 through 3 while the bottom plot is pixels 4 through 6. You see a strong
signal around sample 800 (when the amps were installed). A smaller signal
occurs for pixel 2 around sample 600. This was when the amps were first
installed in only pixel 2. The pixel 0 show more compression than the other
pixels.
-
The next three links plot the compression versus azimuth, zenith,
and julian date.
-
All
the data (.ps) (.pdf)
: This has all the data from 01feb05 thru 30apr05. The dashed vertical
lines in the bottom plot show when the amps were installed. The az,za dependence
is a bit misleading. The gap between az=50-90 is because there was not
data taken here while the amps were in.
-
When
the amps where in (.ps) (.pdf):
This data is for the three days when all the amps were in (and
no attenuators were installed upstairs). It seems strongest for azimuths
around 110 and 290. This is about 180 degrees apart. It may have something
to do with the orientation of the azimuth arm and the aerostat direction
( an azimuth of about 240 deg).
-
When
the amps where out (.ps) (.pdf):
These plots are for when none of the amps were in. There is no clear azimuth
dependence. It may be a little stronger at higher za.
Summary:
-
The amps caused alfa to go into compression a large fraction of the time
when the aerostat was up.
-
Pixel 0 seems to compress more than the other pixels. This probably has
nothing to do with the gain of pixel 0 since the aerostat signal is coming
in through the sidelobes.
-
The attenuators (installed 12apr05) did not seem to affect the compression
on the 13th (the aerostat was only on for the latter part of the run).
On the 15th of april the aerostat was on for the run but there was little
compression (maybe the attenuators were installed on the 13th??).
Previous reports of this problem:
This compression was previously reported in:
processing: usr/p2030/rfirdr/aero/pltaero.pro
home_~phil