The 174Khz ripple in pixel 4A
what we know about the ripple
Processing the data to see the ripple.
Tests show that ripple is in pixel4A fiber receiver.
How the ripple changes over time.
Pixel 4a Ripple goes away after we
the pixel4a fiber receiver.
Other pages related to 174Khz ripple:
ripple: first seen sep04
ripple: reported by alfalfa may05
A ripple in beam 4a of alfa was found back in sep04
while working on the galfa spectrometer (pixel4a
ripple sep04). It looked like it was coming from the fiber optics
Electronics replaced the fiber connectors (some time in oct04??). This
did not make the ripple go away.
The ripple was reported again by a2010 alfalfa group
may05). They were seeing it in the wapp data.
On Jun05 galfa group a2011 again complained about
A timeline of when we did the cleaning/tests:
29jun05: cleaned the fiber rcvr input to pixel 4a. The ripple got
but the next day it was bad again.
30jun05: We swapped the fiber inputs 4a<->4b and the problem
in 4a so the problem is not in the fiber, it is in the fiber receiver
01jul05: opened the fiber receiver box. There is a feed barrel where
fiber enters. We cleaned the fiber on the inside of the box. This did
improve the ripple.
01jul05: We noticed that plugging the fiber into the downstairs fiber
is extremely sensitive. We disconnected the fiber and then reconnected
it (to look inside the fiber receiver box). After reconnecting it,
was no fail light on the fiber receiver front panel but there was also
no signal in the control room. We unplugged and then reinserted the
into the receiver feed through connector and then the signal came
This same problem occurred on 29jun05. Alfa has large gain variations
temperature. It may be that the problem is in these feed through
upstairs and downstairs.
05jul05: we cleaned the upstairs fiber connection (that plugs into the
fiber txmter). The ripple in pixel 4a got worse after we did this.
08jul05: we replaced the fiber receiver for pixel 4a and the
went away. pixel 2a still has ripple (.2% of Tsys).
What we know about the
The ripple is in both the wapps and galfa so it is not a back end
A 174 Khz ripple is a 5.747 usec delay. This is 862 meters in air. The
index of refraction for the fiber cables (platform to clock room)
is about 1.462 (accuracy??). This gives a distance in the fiber
of 590 meters. Eddie castro thought there were 549 meters of fiber
The discrepancy is about 7% (so maybe the index of refraction was not
accurate). The ripple distance points at the fiber as the problem.
For a standing wave to occur in the fiber there needs to be a
at the bottom and and a reflection at the top end. Removing either of
will make the standing wave go away.
The fiber cables for the non alfa iflo are very lossy (I seem to
18 db or more although i don't know whether this is all in the fiber).
The alfa fibers have about a 3db loss. Since the standing wave in
the fiber has to pass through the fiber attenuation twice, alfa fibers
will have a larger standing wave than the older fibers. This also shows
that putting a 3db pad in the fiber will reduce the regular
by 3 db and reduce the ripple by 6 db.
The IF outputs of the downstairs fiber receiver were switched pixel
The ripple moved from 4a to 2a so the problem is before the output of
downstairs fiber optic receiver.
We switched the fiber inputs to the downstairs fiber rcvr
The problem stayed in 4a. This says that the problem is in the fiber
receiver box downstairs (this also says that we are getting a
for both 4a and 4b upstairs...).
The ripple is also seen in other pixels: 0a,1a (1b has been dead so not
sure about this one). 2A, 3A,3B. The strength of the ripple in these
pixels is at least 3db weaker than pixel 4a.
The strength of the ripple is not constant day to day.
To monitor the ripple strength i did the
Processing the data to see
the ripple: (top)
Compute the average spectrum using 200 seconds of data.
Do a robust fit of a 7th order harmonic and a linear polynomial to each
baseline divide the average bandpass by the fit and then
1. The data is now normalized to Tsys
For display purposes, smooth the spectra by 19 channels. the ripple of
interest has a period of 39 channels.
Compute the transform of #2 y=abs(fft(spc))*2. Channel 39
will be the delay for the 174 khz ripple. The amplitude will be in
We did a series of tests with the fiber connections
downstairs to see where the ripple was coming from. The picture shows
3 tests that we performed.
Tests that show that the ripple is
in pixel4a fiber receiver. (top)
from the platform enter on the left and arrive at the fiber rcvr
There is a bulkhead barrel connector that mates the fiber from the
with the fiber that goes to the receiver.
The 3 tests were:
To perform the tests we had to connect and disconnect the fiber
into the chassis and the fibers inside the chassis. Some of the change
in amplitude could be from these changes. The important result is that
there was never a standing wave in pixel 4B. So the problem is
associated with pixel4A receiver and/or interior fiber.
Normal configuration. Pixel4A xmter goes to pixel4A receiver. Test
1 has the ripple in pixel4A (.ps) (.pdf).
This data was taken on 050630.
Pix4a xmter goes to barrel4B and then to pix4B receiver. Test
2 still has the ripple in pixel4A (.ps) (.pdf)..
This means that the problem is in pix4A rcvr or the pix4A barrel. It
tells us that there is no reflection in pix4Brcvr (since we know from 1
that pix4Atx has a reflection). This data was taken on 050630.
Pix4a xmter goes to barrel4A and then to pix4B receiver. For test3
the ripple remained in pixel4A (.ps) (.pdf).
This rules out pix4A barrel. So the problem is in the pix4A fiber optic
receiver (or the small cable that goes from the barrel to the rcvr).
data was taken on 050705.
The first set of plots shows the ripple
on 28jun05 (.ps) (.pdf)
(galfa a2011 file .0002).
Page1: These are the spectra for polA pixel 0a through pixel 6a.
The vertical axis units are Tsys. The high frequency ripple in 4a is
174 khz ripple. It is about .5% of Tsys. You also see a little of the
in pixel 3a. The slower ripple is the 1 Mhz ripple from the dish (and
residuals from the 7th order harmonic fit).
Page 2: Same plots for polB.
The variation of the ripple strength is seen
in the a2011 data taken 20may05 through 29jun05. For each day that they
observed, 1 file was processed and the amplitude of the 174Khz ripple
computed. The plots show
the ripple strength versus time (.ps) (.pdf).
You can see that there are large changes in the ripple strength.
Page 1 : The amplitude of the bandpass ripples (abs(fft(spc))*2)
for ripple number 0 through 100 (the 174 Khz ripple falls at 39). The
ripple if beam 4a. You also see ripples in beams 0,1,2,3 .
Page 2: The strength of ripple number 39 (the 174Khz ripple)
day. The top plot is polA and the bottom plot is polB. Each color is a
On 08jul05 we replaced the fiber receiver of pixel
with a spare fiber receiver. The ripple went away. We then ran 19 200
integrations and computed the strength of the ripples for all of the
The plots show the
bandpass ripples after pixel 4a fiber receiver was replaced (.ps) (.pdf)
08jul05: Pixel 4a
goes away after we replace the pixel4a fiber receiver: (top)
processing: x101/050630/a2011ripple.pro,fibtest.pro, 050708/chkfiber.pro
The strength of the bandpass ripples are plotted versus the
of cycles (of the ripple) that fit in the bandpass (i've limited it to
100 cycles max). Black in polA and red is polB. There are 19 200
integrations over plotted on each frame. The strength is in units of
Pixel 4a no longer has any measurable ripple (it had been .005
Pixel 2A still has ripple at the .002 Tsys level. We will try to get
other spare fiber receiver working and see if we can get rid of it.