lbw check the rf filterbank
11mar06: check filterbank
22apr13: check lbw filter bank.
(Barbara) reported instabilities in polB of lbw
during apr13 runs. They were using the 1230-1470 filter with
the 1370 hipass filter.
On 10apr13 we did a cable test on the 1370
hipass cable going into,out of the filter bank. We saw no
jumps in the signal. A loose feed through connector was found
at the output of the filter bank (it was tightened). A2749
problems continued after this.
On 22apr13 a test was done to see if the rf
switches in the filter bank were making lousy contact.
The test consisted of:
- telescope sitting at az=294, dome=10.9444 degrees.
- put the 1 second toggle into port 7 of the cal
- Use the interim correlator. 4 boards 12.5 MHz bw. take 20
seconds of data
- with the winking cal this gave 10 sets of calon,caloff
(1 sec each).
- for each of the 9 filters:
- switch to another filter then back to the filter
- take 20 seconds of data
- repeat this 10 times.
- then move to the next filter.
- The idea is that sometimes the switch may not work
correctly and we would see a jump in the levels.
- For most of the filters, the 1370 hipass filter was left
in front of the filter bank. This was to get rid of the
- this was not done for the filters whose entire band sits
below 1370 MHz.
- For most of the filters the frequency was set to 1405 MHz
(except those filters that did not cover this frequency).
- Note: the 1280 hipass filter was not included in the set
The data processing was:
- For each filter compute:
- compute the total power for the cal
- normalize the 10onoffs * 10 loops to the median value of
- If one of the filter switches gave a bad connection,
then the cal deflection would show a jump at the start of
the 20 secs of data.
The plots show the
calDeflections from the test (.ps) (.pdf
- Page 1: total power cal deflection:
- Top Frame: polA, bottom frame PolB
- The Colors show different filters being tested.
- the horizontal axis has the 10*10 measurements of the
caldeflection made for each filter.
- The 4 sbc of the interim correlator are overplotted
(they were all set to the same frequency).
- The vertical dashed lines show where the filter was
switched away from and then back to the filter being
- Page 2: cal off total power
- The total power for each cal off was plotted.
- The large variation in the cal deflection for the
1100-1800 filter was caused by a source drifting through
- The blue 1550-1820 filter (blue) had lots of rfi in the
band (at 1700 MHz).
- no jumps were seen in the filter bank switches to the 1-2%
- The test assumes that the switch problem is intermittent.
If it remains for each of the 10 cycles of a switch then the
20 seconds of data is not enough time to see anything drift
back to the normal value.
11mar06: check lbw filterbank
A diagnostic was
written to test the lband filterbank.
The data taking consists of:
- Set the cal to hcorcal and the cal port to the 1 second
winking cal (external port 7 with the 1 second toggle plugged
- For each of the 9 filters banks: select the filter and the
then take 20 seconds of data (10 with cal on, 10 with cal off).
- Do this for each of the 9 filters. Then loop back and repeat
the sequence 9 times.
Use the following definitions:
processing the data:
For each filter we have N sets of 10
measurements of calon,caloff (with 2 seconds between each
measurement. We compute:
- Tsys - temperature from sky, omt, ground
- Tcal - temperature from cal diode
- Ttm - time dependent Temperature caused by the filter
switches not closing properly or changes caused by sources
drifting through the beam. The changes in Tsys between the
calOn,calOff will probably be small (spaced by 1 second).
- Gfilt - gain of the filter. This could
change if the switches did not close properly
- Gif - gain for the rest of the IF.
This will change as we adjust the attenuators.
The calDeflection will
show changes caused by Gif or Gfilt. It will not be sensitive to
changes in Ttm (at least those that change slower than 2 seconds).
- CalDeflection=(calOn-calOff) = (Tsys+Tcal+Ttm - (Tsys
- CalRatio = CalDeflection/Caloff =
- For each of these we compute the median spectra over the 10
The calRatio will show variations in
Tsys caused by Ttm. Ttm can change if the filter relays do not
close properly and increase Tsys. It will also change when sources
drift through the beam increasing Tsys. When sources drift
through, it should occur in both PolA and polB (since most sources
are not highly polarized). The calratio will not be sensitive to
changes in the gains.
- 09mar06. 6 complete passes were made through the
filters. The plots show:
the 9 filters (.ps) (.pdf):
each filter there is a separate plot for polA and polB. The
median calRatio spectra (over the 10 calonoffs) is plotted.
There are 6 in each plot since we made 6 complete passes
through the 9 filters (each spectra is separated by about
9*22=200 seconds). The y axis has units of Tcal/Tsys. If a
filter did not close properly then you would expect the entire
100 MHz band to be affected. There are some jumps in
individual 25 MHz bands. These are probably caused by the
digitizer for this band being hit hard by rfi.
the 6 filters measured at 1370-1470 MHz (.ps) (.pdf):
filters were measured at 1370-1470 MHz. They are over plotted
here to show the calratio measured with different filters.
There are 6filters*6 passes over plotted in each plot.
(1280 hipass) shows Tsys variation (.ps) (.pdf):
top figure shows the fractional change in Tsys for polA
(black) and polB (red) for the 10 cal off measurements that
were made the first time the 1280 hipass filter was measured.
Pol A varied by about 10%. PolB was stable so this was not a
continuum source passing through the beam. The bottom plot
shows the cal deflection. This remained relatively stable.
This shows that the temperature was changing rather than a
gain variation (since the cal deflection would have seen a
gain variation). This variation was not seen in the other 5
passes through the 1280 hipass filter. This variation
started when the filter was selected and lasted for the 20
seconds of the measurement. It did not show up in the Calratio
plots since the median value of the 10 measurements is close
to the normal value. Looking at the spectra, the change was
across the entire bandpass.
- The filters do not show any variation in Tsys except for
the first measurement of the 1280 hipass filter. On 10mar06
this was repeated and no variation in filter 9 was seen.