Ripple in pixel 4A
Alfa pixel 4 polA has a ripple in the bandpass with
a period of 174 kHz. It was discovered while debugging the galfa spectrometer.
The problem was initially blamed on a bad board in the galfa spectrometer.
The same ripple was seen using the wapp spectrometer so the galfa boards
are fine (or at least they don't create this ripple...). The plots show
data taken on 03sep04 (proj a1946 drift scan) using the wapps and the galfa
spectrometer. The wapp took data for 15 minutes. galfa took data for 5
minutes starting 48 seconds after the wapp start. The ripple is seen in
ripples seen by the wapps (ps) (pdf)
ripples seen by the galfa spectrometer (ps) (pdf)
Figure 1. This is the average of 15 minutes of data taken with 100 Mhz
bandwidth and 4096 channels (25 khz resolution). The first lo was
set to 1635 Mhz. The red is polA and the green is polB. The plots are labeled
B1 thru B7 this corresponds to Pixel0 thru pixel6. The ripple is in pixel
4A (B5 red trace). You can not see anything on this scale.
Fig 2. I fit a 3 order polynomial and a 12th order sin/cosine to the bandpass.
The fit was subtracted from the data and then this difference was divided
by the mean value of the fit (over the "good" channels). This put
the data in units of Tsys. The horizontal scale was expanded to 10
Mhz. You can now see the ripples in pix4a (B5 red trace).
Fig3. Pixel4a was plotted by itself using 3 Mhz of data. The oscillation
has a period of 174 Khz. The amplitude is about .005 Tsys (a peak to peak
of about .01 Tsys).
On 08sep04 we switched the IF input cables to the distribution
amplifiers for pixel 2a and pixel 4a. After doing this the ripple moved
from correlator board 5 (pix4) to correlator board 3 (pix2). So the oscillation
is occurring before the IF distribution amplifiers in the receiver room.
Fig 1. This is the average of 5 minutes of data. It started 48 seconds
after the start of the wapp data above. The horizontal scale is in galfa
channels (872 Hz/channel). The amplitude is in galfa counts. The bad pixel
is in B5 red. You can see a little of the oscillation on this scale.
Fig 2. This is pixel 4 (B5). A baseline has been removed (2nd order polynomial
and 6th order harmonic) and then smoothed to 38 Khz resolution (about the
same resolution as the wapp data). The baseline was removed and then the
difference spectrum was divided by the mean of the good channels to put
the difference spectrum in Tsys units. Three Mhz of data was then plotted.
The frequency scale was computed from the 1st Lo (recorded in the wapp
data) and assuming the 2nd lo was 256.25*2 Mhz. I tried lining up the 60
MHz drifting harmonics in both data sets. They were off by about 2 Mhz
(or 62 Mhz) so my assumption about the 2nd lo may have been wrong). The
amplitude of the ripple is .04 Tsys (.08 Tsys peak to peak). This is 8
times stronger than the wapp data. The only difference is that the wapp
data was a 15 minute average while the galfa data was 5 minutes (or that
the frequencies plotted did not overlap).
A ripple of 174 Khz corresponds to a one way distance
of (300/2 meters * 1./.174 usecs) = 862 meters in air. The index
of refraction of the analog fibers that bring the signal down from alfa
is 1.462. A 174 Khz ripple in this type of material would have a distance
of (862m/1.462)=590 Meters. Eddie castro thinks there is about 1800
feet (549 meters). The difference (590-549) is about a 7% error. This could
be in the measurement of the period, distance, or index of refraction.
There is a 174 Khz ripple in pixel 4a of the alfa data.
It is seen in both the wapp and galfa spectrometer.
It has been there since at least 21aug04 (that is the first data I had
to look at).
It is occurring before the distribution amplifiers in the control/receiver
The most likely culprit is a reflection in the fiber optic cable for pol4a.
We need to clean the ends of the fiber an make sure that the connection