Intro
The polB 20db xband hub amplifier had a
problem with the power connection. Both of the 20db xband hub
amplifiers were replaced on
29jun21 with 2 11.5 db amplifiers. The short rf cables were
added to the input so the input cables were no longer
stretched.
On 01jul21 the xband receiver was monitored
for 17hours. The setup was:
- data taken: 16:28 01jul21 to 09:21 02jul21 AST
- telescope sitting at Az=1 el=75.
- xband receiver using 8050 to 9200 MHz
- Mock spectrometer used to record the data.
- 7 172 MHz bands centered at:
- 8132,8296,8460,8624,8788,8952,9116
- Each band spaced by 164 MHz
- 2 pols, 4096 channels/pol, 42 KHz channel width
- spectra dumped once a second
- The temperature sensor in the pedestal was offline during
these measurements.
This page looks at the spectral density function for the 16
hours. The monitoring of the total power can be found here.
The original intent was to look at rfi for the 16 hours. The
ripples in the band pass made it hard to be very sensitive.
Instead, we looked at the standing wave.
- The 1 second spectra were first input.
- the dc spike in the middle was interpolated across using
the adjacent channels
- For each of the 16 hours:
- The average spectra was computed as well as the rms/Mean
by frequency channel for the 3600 1 second spectra.
- for the hourly average spectra and the rms/mean by
hour, the 7 172 MHz bands (separated by 164MHz) were
interpolated to a single band of 42 kHz spacing covering
975 MHz.
- For each second of the 16 hours:
- The 7 spectral bands were interpolated to a single
spectra
- the acf was computed.
- a gauss fit of the peak in the acf was then performed to
track the amplitude and lag of the peak (standing wave
location).
Plotting the spectra.
The first plots show the average
spectra (.ps) (.pdf) :
- Top: spectra averaged for 16 hours.
- black polA, red polB
- This is after interpolating the 7 bands to 1 spectra.
- middle: Spectra by hour normalized to 16 hour average.
PolA
- Each color is a different hour of data
- The bottom is 16:00, the the is 08:00
- an offset has been added for display.
- You can see a ripple that decreases into the night and
the increases again.
- The third band seems to have a small offset. Maybe
the levels are changing during the hour?
- the values seem to diverge from a straight line at
higher frequencies.
- bottom: spectra by hour polB
- the ripple in the spectra is also visible.
- the normalized spectra remain flat across
frequency.
Plotting the rms/mean by frequency channel
For each hour of data (3600 seconds), the
rms/mean by freq channel was computed for each 172 MHz band.
The 7 bands were then interpolated to a single rms/mean by
channel covering 975 MHz.
The plots show the rms/mean by channel for the 16 1
hour data sets.
- Each frame of the plot has the 16 rms/mean for a 200 MHz
span.
- There are 5 frames to cover the 975 MHz.
- Each color is a different hours worth of data.
- For both pols, the ripple spacing is about 13.5 MHz
- The ripple was not present for all hours.
rms/mean
by channel for polA (.ps) (.pdf)
- PolA had a few hours where the rms/mean drifted to
higher values across frequency.
- there are a few hours where the value jumps vs
frequency.
rms/mean
by channel for polB (.ps) (.pdf)
- PolB also showed some hours where the value jumped vs
frequency.
Tracking the location of the ripple
with the acf
A sine wave in the spectra is a
spike in the acf (autocorrelation function). To track the
location of the peak in the acf:
- The average spectra was computed for the 16 hours, in each
of the 7 freq bands
- For each second of the 16 hours
- normalize the 7 spectra to the average value
- interpolate to get a single spectra of 975 MHz.
- compute the acf of the 975 MHz spectra (i didn't bother
to zero extend the spectra so i lost a little resolution).
- The time resolution of the acf was 1/975 MHz or 1.026
nanoseconds.
- Using 51 lags about the ripple peak (lag 71) fit a
Gaussian to the peak.
The plots show the results of the gauss fit to the
acf vs hour of day (.ps) (.pdf)
- Black is polA, red is polB.
- Top: amplitude of the fit.
- The amp is strongest in the late afternoon (when it was
probably hottest)
- Pol B red has a continuous plot
- Pol A has a jump around 19:00 hours. This is also seen
in the total power vs time around
19:00 hours
- The noise around 20 hours is probably caused by the
fit failing (the peak got too small).
- Middle: the location of the peak in the ACF (in micros
seconds)
- the median value was about .0735 (A), .0733(B)
microseconds. If the peak is a reflection with vel=c, then
the dist stance is 150*[.0735,.0733] = 11.02, 10.99 meters
- Bottom: the width of the Gaussian fit.
- for polA the width goes negative around 20 hours (the
fit didn't work).
Summary:
- PolA showed an increase in power with freq for a few
hours. polB did not show this. PolA also had more jumps than
polB.
- There is a ripple in the spectra of both polA and polB.
- The amplitude of the ripple changes with time. It is
larger in the late afternoon when the temperature is higher.
- PolA ripple varies and jumps while polB's is more stable.
- If the ripple is coming from a reflection in the system:
- the average time is .0734 usecs. using 150M/usecs to is
about 11 meters or 36.1feet
- if the velocity in the cable is .7 c, this gives 7.7
meters or 25.3 feet.
- The variation in time of the ripple amplitude is probably
caused by temperature.
- A loose connector could change it's reflection coef with
temp.
- We could find the ripple location by added attenuators at
different points and see how the ripple changes.
- had 2 sharp jumps where the amplitude was a fun
processing: x101/210701/xbspcrfi.pro