Occurrence of rfi versus frequency by month
last month added: aug20
Histograms of interference versus frequency are
computed
for the data acquired on the telescope using the correlator. For
each
scan
of N separate integrations (typically 1 second) the rms/Mean by
channel
is computed. A linear baseline is used to remove any zenith angle
dependence.
Histograms are then made of the number
of
rfiCnts/TotalCnts
by frequency channel. Any rms/Mean greater than 3 times the
expected
sigma
is called rfi. This sigma is computed using the bandwidth,
integration
time, and corrections for: 3/9 level sampling, hanning smoothing,
and
the
instrumental bandwidth of the channels. Any band passes that have
more
than 80% of their channels "bad" are not included in the rfi or
total
counts.
The histograms have .1 Mhz resolution. The rfi count in the .1 Mhz
channel
is incremented by 1 (and only 1) if any of the telescope channels
within
the .1 Mhz bin exceed 3 sigma.
There is a cumulative plot for the year
followed
by monthly plots. Each plot has the fractional occurrence of rfi
in
black.
The blue histogram is the total number of counts (nonrfi + rfi) in
each
bin. The vertical scale on the right tells the total counts per
bin.
The
red line is a histogram of the number of times a bandpass was
rejected
because more than 80% of the frequency bins were above 3 sigma.
This
histogram
has been scaled to: badcnts/(total counts + badcounts). If there
are a
large fraction of red lines it may be because of some mapping
programs
that include large continuum sources or have large za variations
for
the
strip (some of the alfa proposals).
PLOTS OF rfi by year and month.
The entries without links have no data for that month. The 611 and
327
receiver have not been installed a large fraction of the time.
Occasionally
the rfi monitoring will take data on the receiver when it is not
there
(so there is just a load connected). This depends on whether I
remember
to update the receiver file to use (and that electronics reminds me
that
they have put it up or down). Sband 1800-3100 switched from just
sband
narrow (2320-2420) to the full range of 1800 to 3100 sept01 when the
sband
wide receiver was reinstalled. The pattern of dips in the total
counts
by bin (blue lines) is from the x111 datataking. It steps by 23 Mhz
giving
an overlap of 1 Mhz and the plotting ignores 1 Mhz from the edge (do
to
filter rolloff). If you use ghostview to look at the plots, clicking
on
a mouse button will blowup the region centered on the mouse.
Why to do the plots look lousy :
- The plots are washed out with linux gv 3.5.8??
- Turn off anti-aliasing (click on state, disable antialias)
- I don't see all of the plot:
- linux gv 3.5.8 .. If it says BBOx (bounding box) at the top,
click on
it
and set it to letter. Push the cursor button and pan around to
see
parts
of the plot that are not visible.
- kghostview (fedora) click on view -> paper size..
select
anything
but auto (letter, or a4, or whatever..)
Year 2011 comments:
- 430 Mhz rfi below 418 Mhz.. ignore since filter goes 418 to
422.
Year 2010 comments:
- mar10: x111 was run during the telescope beam repair. punta
salinas was running in all modes during this time so lband looks
a lot
messier.
Year 2009 comments:
- apr09: included option to ignore certain alfa beams when
making
histograms. Beam 6b bad was causing the slope in alfa data from
1350 to
1400
Year 2008 comments
- sep08: alfa has decreasing slope of rfi 1350 ->1420... This
is
being caused by beam 6b which is dead. it's noise changes across
the
band.
- jun08: sbh,band 3450 to 4200.. pretty messy. This was a2355
doing
300 second position switching with source that had continuum.
Not sure
why it looks so bad.
Year 2005 comments
mar05 - The rms for the wapp data was computed and then stored in
the
old
interim correlator format back out on disc. The frequency flip was
occasionally
getting screwed up. Fixed the problem and then processesed the
data
nov04
though feb05. The lbw/alfa birdies should now be in the correct
positions.
feb05 - updated exclude criteria for wapps. Some patterns
were
being
included that shouldn't have been. Reprocessed the data.
Year 2004 comments
nov04 - include alfa data and wapp data for the other recievers.
feb04 - 6 to 8 Ghz (cbh) installed.
Year 2003 comments
sep03 - xband. the blocks of 25 Mhz that have rejected bands (red)
is
from
the lousy pointing at xband. Some users were doing on/offs on
strong
continuum
sources. The power in the on varied over the scan. When computing
the
rms
by channel you got large values. Since the continuum source affect
all
channels, the band got rejected (red gt 85% of channels bad)..
feb03 - telescope shutdown. drift scans 327,cband (4800) , and
lbndrift.
feb03 - sbwide,lbw switch in filters while stepping. especially
sbwide..
Year 2002 comments:
sep02 327 data was being taken at the wrong frequency. Was using
260IF
and trying to get 4 sub bands out of it. When the new
receiver/filters
were installed in jul2002, the file x111auto.dat was not updated
for
the
new bandwidths. Was corrected on 16sep02.
aug02 327
2
Mhz comb from 430 waveguide power monitor.
Year 2001 comments:
jan-mar 1366,1382 radar (more
info)
March: seti observations, not much data.
April: Tv station at 1408 (more
info)
May: Tv station at 1408 (see above link)
July: aeronomy runs, not much data.
August: A1459 baseline instabilities (1340-1440) from air
conditioner
in dome (more
info).
Sept: A1459 baseline ripples (see above).
sband wide. The Yr2001 data shows that the fraction of band passes
that
were rejected (the red lines) stays high up to 2600Mhz and then
drops
to
zero. I think this is because the birdies at 1940 Mhz are
overloading
some components after the 1 Ghz filter of the 10 Ghz upconverter.
When
you are farther than 500 Mhz from 1940, the cellular phones at
1940
start
to get filtered out. The culprit is probably the fiber optic link.
When
the power levels are adjusted, the total power is used. It is high
from
the 1940 birdie so the spectral density for the rest of the band
gets
set to low.
processing: rfi/rms/mkyr.pro
home_~phil