Polarized Tsys using the alfa receiver
sept 2005
Links:
Rotating the azimuth arm (section):
(TsysATsysB)
vs az for each pixel and za (.ps) (.pdf):
2az
amplitude and phase of maximum (.ps) (.pdf)
Rotating the alfa feed system (section):
TsysK,
TsysDifK, and the fits versus rotation angle (.ps) (.pdf
) :
Introduction:
Polarized components of Tsys have been previously
measured with the lbw
, sbh,
and cband
receivers. These tests were done by rotating the azimuth arm and
measuring TsysATsysB versus azimuth. A sine wave of period 180 degrees
and amplitude .2 to .5 kelvins was seen. The phase of the sine wave depended
on the orientation of the feed probes (which differed for each feed).
The alfa receiver has 7 feeds. It also has the capability
to rotate +/ 90 degrees about the center pixel. All of the receiver probes
have the same alignment direction. When the alfa rotator is at 0 degrees,
the two probes (polA,polB) point +/ 45 degrees from the azimuth arm direction.
Two sets of tests were done with the alfa receiver:

22aug05: TsysATsysB vs azimuth at za 5.7.5,10,12.5,15,17.5, and 19.5 degrees.

01jun05: TsysA  TsysB vs alfa rotation angle (with the telescope sitting
at az=270 degrees). Rotate 90 to +90 and back two times.
22aug05 TsysATsysB vs azimuth
angle: (top)
TsysATsysB was measured versus za using the following
setup:

The alfa rotator was set at 0 degrees while the telescope was moved from
.

The telescope was driven in az at .4 deg/sec from az 270 to 630 (or
630 to 270). This was done at zenith angles of 5,10,15,19.5,17.5,12.5,
and 7.5 degrees.

The wapps were used to record the data using a 100 Mhz bandwidth centered
at 1440 Mhz. The data was sampled once a second.

The cal was fired 4 times during the measurements. The average values was
computed (there was little variation) and then used to convert the data
to degrees K.

This was done during the afternoon so the sun was a source of interference
when a sidelobe of the feed was close to it.
The data analysis was then:
For each az swing sample compute the total power over the 100 Mhz bandpass
(throwing out outliers).
Scale the data to kelvins using the cals.
Compute TsysPolAK TsysPolBK for all beams.
For each az swing fit yfit(az)=c0 +c1*az + c2*cos(az)+c3*sin(az) + c4*cos(2az)+c5*sin(2az).
The routine corblauto() was used for the fitting. It iterates the fit throwing
out points greater than 3 sigma.
The plots show the difference (TsysATsysB)
vs az for each pixel and za (.ps) (.pdf):

Each page has (TsysATsysB) vs az for the 7 pixels at a single zenith angle.

The black line is the data (TsysATsysB). The red line is the fit.

Fig 1. za=5. degrees.

Fig 2. za=7.5 degrees. Large interference az 90 to + 30 degrees.

Fig 3. za=10

Fig 4. za=12.5 . Interference 90 to 0 degrees.

Fig 5 za=15.

Fig 6 za=17.5 . Interference 90 to 30 degrees.

Fig 7 za=19.5. Interference 90 to 20 degrees.
The azimuth swings were done in the sequence: za= 5,10,15,19.5,17.5,12.5,7.5.
The interference was probably caused by the sun.
The coefficients of the 2az terms were used to compute
the amplitude and the azimuth angle of the sine wave maximum. The 2az
amplitude and phase of maximum (.ps) (.pdf)
are plotted versus za.

Top: The 2 az amplitude vs za. Each color is a different pixel. The scatter
in za=7.5 and 12.5 is probably caused by the interference in those measurements.
The mean value is about .29 Kelvins.

Bottom: The azimuth angle for the maximum of the 2 azimuth term. The median
is 47 degrees.
Summary:

When polA measures the maximum of the 2az cycle, polB is located 90 Degrees
away at the minimum. TsysATsysB is twice the amplitude. So the polarized
component of Tsys is .3/2 = .15 kelvins.

The panels on the dish are aligned with the long direction of the panel
sitting east/west. The data was taken with the alfa rotation angle set
to zero degrees. The alfa probes are then aligned with az45 and az + 45.
When az=45 degrees polA is pointing along the long direction of the
panel (az=90) and polB is aligned with the short direction of the
panel (az=0) (it's possible that polA,B are reversed here). This would
give the maximum difference in the polarized component of Tsy if
the polarized power is coming from the panels. The measured value for the
maximum of az=47 is very close considering the interference. So it looks
like the polarized power can be attributed to the panels on the dish.
processing: x101/050822/azswing_fit.pro, azswing_inp.pro
01jun05 rotating the alfa feed.
(top)
TsysATsysB was measured versus alfa rotation angle
on 01jun05. The azimuth arm was stationary at 270 degrees (dome in
the west) and the zenith angle was at 8.468. The rotator was moved from
90 to 90 in 100 seconds. 4 rotations were done (90 to 90,90 to 90,90
to 90, and 90 to 90). The wapps were run in pulsar mode with 100 Mhz bandwidth,
256 lags, centered at 1420 Mhz. The spectra were sampled at a1 millisecond
rate. The cals were cycled at a 25 hz rate (20 milliseconds on, 20 milliseconds
off) synchronized with the 1 second tick.
The processing was:

Compute the 100 Mhz total power for each millisecond.

Compute the median value for each 20 millisecond cal on or cal off. Throw
out the 1st and last millisecond of each 20 millisecond set (for cal transitions).

Compute TsysK=(calOff/(calOncalOff)) *CalValueK where calValueK
is the cal value for each pixel at 1420 Mhz.

Compute TsysDifK= TsysKATsysKB for each pixel and sample.

Average the 4 sets of TsysDifK.

Fit yfit(th)=C0 + C1*th + c2*sin(2*th + C3) to TsysDifK (here th
is in radians).
The Pixel positions: 90 degrees start and +90 degree end are shown in
the picture below. They are sitting on the feed array with up towards the
secondary (and pointing west since az=270). The rotation was clockwise
looking at these pictures.
The plots TsysK,
TsysDifK, and the fits versus rotation angle (.ps) (.pdf
) :

Fig 1: Tsys vs rotation angle. Black is polA, red is polB. The 4
rotations have been over plotted. The structure is repeatable for the 4
passes. This shows that the changes are not coming from the sky (since
the sky moved during the 400+ seconds it took to do these measurements).
The spikes (negative and positive) are spaced in time by 12 seconds. They
are coming from the FAA radar at 1350 Mhz.

Fig 2: TsysATsysB was computed for each of the 4 rotations and
then averaged. The black lines are the averaged data. The red line is the
fit to the data. The fit used does a pretty good job of fitting the data:
linear in theta, and 2*theta.

Fig 3: The coefficients from the fits to each pixel are over plotted.
The x axis is the fit coefficient index:

C0. This is the constant TsysATsysB (in degK). This probably comes from
the cals being incorrect.

C1.*theta. The linear term in the rotation angle theta (in degK). Each
pixel is plotted in a different color.The fit was done in radians so theta*C1
would go +/ .6*1.5= +/ .9 K. It is interesting to see how the C1
value groups by pixel: pixels (0,2,5)=0, pixels(3,4)
increase with rotation, pixels (1,6) decease with rotation. As if
there was a linear gradient in polApolB that increases left to right
in the above picture.

C2*sin(2theta) This is the amplitude of the 2*theta term (in degK).

C3 where sin(2theta+c3). The phase of the 2theta term (in Radians).

Fig 4 The coefs are plotted by pixel.

C0 The constant difference (TsysATsysB) does not look that random when
plotted versus pixel number. Maybe there was a systematic error in the
measurements of the cals.

C1 The linear term. Looking at pixel 1 to 6 there is a cosine term with
period 180 degrees as you move around the 6 outer pixels.

C2 sin(2theta) amplitude. Pixel 0 is strongest

The rotation angle where the 2*theta term is a maximum (th=(90C3)/2).
It is centered on theta=45 degrees. When theta is 45, the right probe
(looking uphill from the center of the feed system) is aligned with the
azimuth arm. Since the az was parked at 270 degree this makes the maxium
in the east/west direction. This is the save value that was found with
the azimuth rotation. So the maximum is attained when one probe points
along the long axis of a panel (east/west) and the other probe points along
the short direction of the panels (north south).
summary:

The alfa rotation also saw a TsysATsysB sine wave with a 180 degree
period. It was aligned with the east west direction of the panels.
So it doesn't matter whether we rotate the azimuth leaving the feed angle
fixed or rotate the feed and leave the azimuth fixed. In both cases the
2*theta maximum aligns with the panel direction.

The amplitude of the 180 degree period is 2 to 3 times stronger than what
is measured when swinging the azimuth arm. The difference may be
that the alfa rotation is moving in the focal plane while the azimuth rotation
remained fixed in the focal plane.

When plotting the coef's versus pixel number, there are some interesting
groupings and curves. They may be real , or just an artifact of the fitting.
processing: x101/050601/rotalfa.pro
home_~phil