The problem:

    Dynamic spectra of beam 1a:

• 100 Mhz bw, bandpass correct with median spectra (.gif):
• The image has been band pass corrected using the median bandpass over the 600 spectra.
• There are 2 continuum sources (at 120 and 520 secs). They take about 13 seconds to drift through the beam.
• The faa radar is at 1350 with an IF harmonic at 1380 and 1410 Mhz.
• 1381 Mhz is GPS.
• You can see the jumps in the power level that are faster than the continuum sources. These jumps are relatively flat in frequency.
• 25 Mhz section of first dynamic spectra (.gif):
• This is a 25 Mhz section of the first image. You can start to see some high frequency noise (e.g. around spectra 65).
• Remove 2nd order baseline from each spectra of the 25 Mhz image (.gif):
• The total power from each spectra has been removed. The high frequency noise is seen at records 65,  270, 460,  and 580.

Plotting the total power jumps and ripple strength:

The plot shows the avg bandpass, total power jumps, and ripple strength (.ps) (.pdf):

• Average bandpass 7 beams: This has been averaged over the 600 seconds. Black is polA and red is polB. The dashed green lines are the portion of the spectrum used to measure the ripple strength.
• Total power vs time for 7 beams: The total power for each spectra was computed (excluding the regions with rfi). The total power time series was  normalized to the median value for each beam and then plotted versus time. Black is polA and red is polB. The bottom trace is beam0 while the top trace is beam6.
• The positive going bumps are continuum sources.
• Pol A and polB should track each other. Beams 2,4,5, and 6 have some variation in the polA/B ratio with time.
• Beams 0A and 1A are noisier than the others. Beam 1A (reported as bad) is the worse.
• Rms/Mean for each frequency channel: The rms/mean was computed for each frequency channel over the 600 time samples. Outliers (continuum sources) were not included in the rms computation.  The green dotted line is the expected rms (=1.2(3lvl)/sqrt(100e6/4096*1.2(sinx/x width)*1.)). The rms vs frequency for beam 1a shows a frequency dependence (not seen in the others beams).
• Blowup of avg bandpass showing ripple: A 1st order polynomial and a 9th order harmonic was removed from the 100 Mhz average bandpass for each beam. 5 Mhz (1360 to 1365 MHz) was then plotted for each beam. Black is polA and red is polB. The y axis is in units of Tsys.
• You can see a high frequency ripple in beams: 1A,2A,3A,3B, and 4A. This is the 174 Khz ripple caused by the reflections in the fiber optics cable. This was seen in 07sep04 in the galfa data and again in 11jul05 in the alfalfa data.
• Back in jun05 pixel had the largest ripple (about .003 tsys). We replaced the fiber optic receiver at the time. The other pixels had ripple on the order of 0 to .002 Tsys.
• Avg ACF showing strength of reflection: For each of the 600 spectra:
• The frequency range 1352 to 1418 Mhz was extracted and a 9th order harmonic was removed.
• Each 1 second spectra was then normalized so that its median value was unity.
• For each 1 second spectra the acf was computed.
• All 600 acf's were averaged and plotted.
•  Each beam is color coded. PolA is always the lower trace of each pair. The x axis is in useconds (lag delay). The y axis is in units of Tsys (with offsets for plotting).
• The birdie at 5.782 useconds is about the time needed for a reflection to go up and back in the fibers.
• Beam1a has the largest ripple (.002 Tsys) but it is not much larger than some of the other pixels. It has the same order of magnitude as the values measured back in jun05.
• Stability of the reflection:  For each 1 second acf:
• the peak value was recorded for the 5.782 usec delay (actually the peak was taken from 5 lags around this value).
• The amplitude of the ripple was then plotted versus time. Pol A is black and polB is red. The vertical scale is in units of Tsys.
• The amplitude of the ripple was stable for all beams except beam1A.
• Beam1a would jump between .001 Tsys an .005 Tsys.
• The ripple should be a function of the power reflecting at each end of the fibers. Since we normalized each 1 second spectra to unity, there should be no power variation in the ripple power (since the total power was forced to be constant). This means that either the reflection coefficient is changing or the optical transmitter if jumping on time periods of a second or more.

Conclusions:

• Extra noise was seen in beam1A of alfa on 01nov06.
• The total power versus time shows beam1A to be worse than the other beams (beam0a is also worse than the others).
• The rms/mean by channel shows a frequency dependence for beam1A. This means the jumps are not constant in frequency.
• The 174Khz ripple caused by reflections in the fibers is present in a few beams. The amplitude is similar to what is was back in 2004 and 2005 (except for beam1a).
• The amplitude of the reflection in beam1a is changing in time from .001 Tsys to .005 Tsys. The jumps happen on orders of a second.
• The jumps are not caused by changes in total power (since each spectra was normalized to unity). This probably means that the fiber optic transmitter is failing. A less likely cause could be that the reflection coefficient is changing on time scales of a second.