Standing waves measured by moving the platform with the tiedowns.
Standing waves occur between the horn and the dish.
They cause ripples in the spectrum with a period of about 1 Mhz. This corresponds
to a delay of 1 usec. Position switching at AO is used to cancel these
standing waves (since they mostly repeat in the on and the off).
On 04jan05 data was taken to measure these standing
waves. The measurements started at 15:42.
The setup and processing:
The data processing was:
lband wide in linear polarization mode
3 contiguous 50 Mhz bands centered at 1400,1450, and 1500 Mhz. Each 50
MHz band had 2048 channels.
Data was dumped once a second using the interim correlator.
60 seconds of data was taken before the tiedowns were moved. This was used
as the bandpass correction.
The tiedowns were then moved from 12.42 inches to 16.42 inches in 243 seconds
(.016 tdinches/second). They were then moved back down to the starting
position. An increasing tiedown position causes the platform to move lower
so the platform was lowered and then raised. The entire motion took about
This processing was done for platform
moving up, down, polA, and polB separately.
Compute the median bandpass for the 60 seconds before the tiedowns were
moved. Divide this into each of the 1 second spectra during the move.
For moving up and moving down average channels 1100 thru 1500 and then
use these 243 numbers to flatten each spectra (i should have subtracted
the values rather then dividing....).
combine the 3 50 Mhz spectra into a contiguous 150 Mhz spectra.
clip the spectra to +/- .02 Tsys. This is to get rid of the galaxy and
the bandpass edges.
0 extend and then compute the fft of each of these 150 Mhz spectra. This
is the Acf.
compute the magnitude abs(acf) and and phase atan(float(acf),imaginary(acf)).
The bandpass correction was taken from the tiedown starting
position and was used for the entire 20 minutes. Any ripples present
in the spectra near this tiedown reference position should be canceled
by the bandpass correction (assuming they are stable in time). When we
move away from the reference position the ripples should shift in phase
and no longer cancel.
The acf will have a spike if the same radiation takes
two paths into the horn. The lag for the time delay of the path difference
will show increased correlation because the same signal is in the immediate
and delayed elements of the multiplier. For a peak to move from 1 acf channel
to the next, the signal path must increase by c*acfShiftLen. For150 Mhz
bandwidth this is about 1 meter. If you have a complex acf, then the phase
at each lag will shift through 2pi radians when the extra path length changes
by 1 wavelength (21 cm).
The images show the magnitude of the acf versus the
offsets from the tiedown starting position.
Images of ripple strength versus tiedown offset
The ripple at 1.04 usecs is strongest. Weaker ripples
spaced in steps of about .1 usec about 1.04 usecs are also present.
These weaker ripples are coming from the motion of the platform since they
appear and disappear as we go farther/closer to the reference position.
The things close to .1 usec are probably trouble with the bandpass correction
since they are always there.
vs delay. polA,B separate (.gif):
The ripples are repeatable when moving up and down, and between polA and
The top two plots are polA, polB when the platform was moving down (td
extension increasing). You see no ripples until 1 inch of td offset.
Ripples then appear with the strongest at 1.04 usecs. This happens in both
polA and polB.
The bottom two plots are when the platform is moving up (td extension decreasing).
The ripples remain visible g until 1 tiedown inch before the starting position.
vs delay averaging polA and B (.gif): This has the
same data as the above plot but polA and polB have been averaged to increase
the signal to noise.
The platform moved down and then back up. The bottom
plot shows that when the platform returned to the starting position (20
minutes later) the original bandpass correction was still canceling the
ripples. This shows that the ripples are relatively stable in time.
To increase the signal to noise acfs were averaged together.
100 seconds of acf magnitude were averaged when the tiedowns were
farthest from the reference position (when the signal was strong) and .30
seconds of data were averaged close to the starting position (to
use as a comparison reference).
Plotting the amplitude and change of phase:
The plots show the strength
of the ripples and the change in phase with tiedown offset (.ps) (.pdf).
Fig 1 Avg strength of ripples vs delay (polA,B separate). The top
plot is platform moving down, the bottom plot is platform moving up. Red
(polA) and blue (polB) are the average of 100 seconds when the platform
was farthest from the reference position. Green is 30 seconds closest to
the reference position. The vertical scale is in units of Tsys )an offset
has been added for display purposes). The width of the spike at 1.04 usecs
is not resolved. With 150 Mhz bw the distance resolution is 1. meter.
The center of the polB 1.04 usec peak is offset from the polA peak by a
fraction of a channel.
Fig 2 Avg strength of ripples vs delay (avg pols and motion).
Top plot has averaged the polarization's. Blue is platform moving
down, red is platform moving up. Green is when we are close to the reference
Bottom plot averages pols and directions: The black line is the
average of the directions and the polarization's when far from the reference
position. The green line is the average when close to the reference position.
The 8 largest peaks have been flagged (in red) and the delay and strength
of each of them is listed in the table on the right. The 1.04 Usec peak
has a strength of .00089 Tsys. The processing divided by a reference spectrum.
Below we will see that we moved the platform by 6 cm which would
shift the sine wave in the spectra by 30%. Dividing a sine wave by
one shifted by 30% would increase the amplitude by about 50% . The
averaged data covered a 15% to 30% shift in the sine wave so the
.001 Tsys may be a little higher than the ripple value without the
Fig 3 change of phase of 1.04 usec peak: To move between channels
of the acf the distance must change by 1 meter. Smaller changes will cause
the ripple in the spectra to shift in frequency. This can be measured by
creating a complex acf and watching the phase of the 1.04 usec delay move.
A distance of 1 lambda will move the phase 2*pi radians. For the
phase to be defined, the amplitude must be non zero. Close the the starting
platform position the bandpass correction cancels the ripples so the amplitude
is close to zero. As we move farther away, the amplitude becomes non zero
and the phase is defined.
Top plot 1.04Usec phase vs td motion (rawDat): The phase for polA,
polB, moving Up, and moving Down are all plotted separately. This is the
raw data before fixing phase jumps of 2pi radians. All 4 of the measurements
track each other.
Bottom plot 1.04Usec phase vs tdmotion (avgPol): PolA and polB have
been averaged and the phase jumps of 2pi were removed. The vertical scale
was switched from radians to cm. Black has the platform moving up, red
is platform moving down. Both directions track each other. The blue line
is a linear fit to platform motion versus tiedown motion. The fit
gave .642 platform motion/tiedown motion (or 1.56 tiedown motion/platform
motion). The slope is not unity because the tiedown/main cables stretch
a little and the tiedown boom rotates. The value of 1.56 is
a little smaller than what was measured (1.73) fitting a years worth
of data (tiedown motion and platform motion from the distomats).
This was all done at a of ten degrees (where the cos(10) = .09).
Fig 4 Phase versus tdmotion for 8 largest peaks: If a peak in the
acf comes from a reflection that goes down to the dish and back, then the
phase of the peak should change as we move the platform. This was seen
in the 1.04 Usec peak. The phase for the 8 largest peaks (see figure
2 bottom plot) were plotted versus tiedown motion to see if they depended
on the distance between the platform and the dish. The top plot has the
platform moving up, the bottom plot shows when the platform was moving
down. Each color is a different peak (the delays are printed on the left
.. .65,.79,.81... usecs). Linear fits were done to each plot.
Fig 5 Slope of phase vs td motion for the 8 peaks: The slope of
the fits from figure 4 are plotted versus delay (peak location). Black
is moving up and red is moving down. Seeing a slope equal to
the 1.04 usec delay means that the peak goes between the dish and the bowl.
0 slope says the the path length is not changing with the platform/bowl
distance (or the amplitude is so small that the phase is not well defined).
Since all peaks are present after dividing by the reference, they must
be changing. They all go back to 0 when we return to the reference so the
change is related to motion rather than time.
.79 usec, .95usec, 1.04 usec, 1.17 usec peaks are probably reflecting between
the bowl and the platform.
.9 usec, 1.25 usec peaks change with platform motion but with a different
.65 usec and .81 usec don't seem to change phase with platform motion.
German cortes provided the following distances. The
distance in usecs is the round trip time for the distance.
|horn to tertiary
|tertiary to secondary
|secondary to primary
|primary to bottom of dome
|horn to edge of tertiary (shortest)
|horn to edge of tertiary (longest)
|diameter of ray dome
|horn to dish
These distance show that the 1.04 usec peak is between
the horn and the reflector.
plot of the distances (.pdf):
Standing waves were seen that changed their strength as we moved the platform.
The strongest peaks were at:
When we returned to the same platform position (20 minutes later) they
returned to their original values (the bandpass correction canceled them).
1.04 , .79, .91, 1.25, .81, .65, .95, 1.17 usecs (ordered strongest to
The 1.04 usec delay corresponds to the distance between the horn
and the dish. The amplitude was about .001 Tsys. It showed a change
of phase with tiedown motion of 1.56 tdMotion/platform motion. This comes
from the stretching of the tiedown/main cables and the rotation of the
td cable boom.
The delays at .79 , .95, and 1.17 usecs moved with the same
rate of phase as the 1.04 usec peak. They are probably bouncing between
the bowl and the platform.
The delays at .91 and 1.25 usecs change phase with a different rate so
than the 1.04 usec peak. They are reflecting off of the platform and something
The delays at .65 and .81 usecs do not change phase as we move the platform.
The standing wave at 1.04 usecs was not resolved with 150 Mhz bw (1 meter
The 1.04 usec standing wave is shifted a bit in polA and polB (by a fraction
of a channel).
For total power on off observing: If the tiedowns move appreciably between
a position switch on and position switch off, then this could introduce
standing waves. This might happen when we move from raining (distomats
not working) to no rain (distomats starting to work). It might also
happen when the dome is at high za and passes in front of a tower causing
the tiedown cables to lose tension.