The standard heiles calibration scans  do 4 strips across a source while the data is sampled with the correlator  at a 1 hz rate. The spectra are normalized by the off source spectra and then converted to kelvins using the cal. Fits are then  done to the stokes parameters (I,Q,U, and V).

    Project T2212 wants to know the gain as a function of frequency (and probably angle from beam center) over a 20 Mhz bandwidth to compute the back scatter gain. The normal calibration scan processing fits 2d gaussians to the the total power (25 Mhz). This was modified to include fits by individual frequency channels.

---

The setup and data taking:

• 128 channels over a 25 Mhz bandwidth  centered at 430 Mhz.
• A 20 Mhz  Rf filter was used between the dewar and the postamps. This limited the available channels to about 100.
• The interim correlator took the data in  3 level stokes mode (auto and cross correlations on the two polarizations).
• Fits were done to stokes I (polA + polB).
• Data was taken on:
• 30jan05 (after the cavity filters were placed in front of the dewar)
• 15jul06 (before the turnstile was changed)
• 22jul06 (after the new turnstile was installed but before the final tuning).
• The data was taken with the radio astronomy receiver. The monoplexor was not installed.
• The point sources used were:
 

Source	Flux Jy@430	# patterns  30jan05	# patterns  22jul06
B0134+329 (3C48)	37.5	11	19
B0340+048	7.8	19	17

---

Processing:

• The 240 (60*4)  spectra of a pattern were divided by the cal off spectra (taken 3 beams away from the source) to remove the bandpass.
• The spectra were scaled to Kelvins using the cal on, cal off measurements. A single cal value at 430 Mhz was used.
• For each of the 4  strips,  gaussians were fit to the main beam and the two sidelobes.
• The sidelobe fits were removed from each strip and then a 2-d gaussian was fit  to the strip data.
• A harmonic fit was also done to the sidelobe height, position, and width.
• The fitting was done to the total power (averaging over the 25 Mhz bandwidth) and then it was repeated  for each of the 128 frequency channels (actually only about 100 channels were used because of the 20 Mhz rf filter).

---

The fit results:

30jan05 (.ps) (.pdf) : This data was taken after the cavity filters were installed in front of the dewar to get rid of the tv station harmonics. There were 11 patterns on B0134+329 and 18 patterns on B0340+048.

• The total power (25 Mhz) strip data. The top plot has the 11 patterns of B0134+329 overplayed. The bottom plot has the 19 patterns of B0340+048. The dashed blue lines are the start of each  of the 4 strips (60 points per strip).  You can see the main beam and the sidelobes. The data is still in correlator units (Tsys is about 1.)
• BandPass and az/za coverage: The top is a typical spectral bandpass. The spectral density falloff at lower frequenceis is being caused by the 20 Mhz rf filter. The dashed red lines are the channels that were actually fit. The channels outside these lines were given the values at the dashed lines. The bump at 426 Mhz is interference. The bottom plot is the az, za position  for each pattern done.
• Gain: The gain in Kelvins/Jy. The top plot is B0134+329 and the center plot is B0340+048. The gain vs frequency is plotted for each pattern. The green mask is the region that was fit. The spikes  are caused by rfi. The bottom plot is the gain fit to the 25 Mhz bandwidth vs zenith angle. The gain offsets in the first two plots are being caused by the za dependence of the gain.
• Tsys: Tsys in Kelvins. The top two plots are versus frequency. The bottom plot is the 25 MHz bandwidth fit vs za. The offsets in Tsys are caused by the za dependence. B0134+329 has a Tsys that is larger than B0340+048. This is probably leakage from the source getting into the baseline.  The difference is about 3 kelvins. The Temperature difference of the two sources is about 300 Kelvins so the leakage may be about 1%.
• BeamWidth: The average beam width ((az+za)/2). The offsets are being caused by the za dependence (spill over).
• Normalized gain, Beamwidth, and Tsys: The  za dependent offsets were removed by normalizing each pattern to its median value (over the channels used).
• Normalized Gain:
• The normalized gains for the different patterns  are repeatable. The fgain for the two sources (whose strengths differ by about a factor of 5) are also the same.
• The gain peaks around 427 Mhz and drops by about 10% as we get out to 440 Mhz. This seems like a lot.  It could be that the illuminated area (probably on the tertiary) is changing with frequency.
• The ripples at about 1 Mhz spacing are real.  They are the standing waves caused by the source power bouncing between the dish and the feed. They remain constant in phase as we track across the dish and between sources. For the phase to move, the reflection distance would have to change by a fraction of 70/2=35 cm. This is  telling us that the reflections are  coming from the dome/za rails (which remain at a fixed distance from the dish as we track in az,za) and not from the fixed platform (whose distance from the dish changes as we move in az,za)
• Normalized beam width: The beam width is falling off with the expected 1/freq  dependence (the green line). It starts to deviate from this around 436 Mhz.
• Normalized Tsys: Tsys is pretty flat so the normalization by the off spectrum is working pretty good.

15jul06 (.ps) (.pdf) : This data was taken before the new turnstile was installed.  There were 19 patterns on B0134+329 and 14 patterns on B0340+048. The data looks pretty much the same as the jan05 data.

22jul06 (.ps) (.pdf) : This data was taken after the new turnstile was installed. It had not yet had its final tuning.  There were 19 patterns on B0134+329 and 17 patterns on B0340+048. The data also looks the same as the previous data.

Comparing jan05,jul06 (.ps) (.pdf):

• Top:  Gain vs frequency. Black is  jan05. Blue is 15jul06 (old turnstile). Red is  22jul06 (new turnstile).
• 2nd plot: Average gain vs frequency for jan05 and jul06. The median value was taken at each frequency channel. The two jul06 datasets before and after the turnstile installation are the same. The gain above 435 Mhz was a few percent higher back in jan05 than it is now in jul06.
• 3rd plot: Beamwidth vs Frequency. The green line is 1/freq.
• Bottom plot: Average beamwidth vs frequency. The median value was taken at each frequency channel. This has remained constant for the 3 data sets.

---

Conclusions:

• The gain was fit  by frequency channel for data taken in jan05 and jul06.
• The gain peaks around 428 Mhz. It falls off by about 10% at 440 Mhz.
• The data taken in jul06 falls off at 440 Mhz a little faster than the jan05 data.
• The ripples in the gain curve are standing waves between the receiver and the dish.
• The beamwidth vs frequency has remained constant from jan05 thru jul06.
• Changing the turnstile did not affect the gain or beamwidth.
• To compute the back scatter gain vs frequency,  the fit equations can be evaluated at each frequency channel  as a function of offset from the beam center.
• The measurements need to be repeated with the monoplexor installed. In this case we can just duplicate the single polarization (to get the  code to work).
• Probably could give a little more thought to the normalization of individual spectra. An average throwing out outliers is probably better than the median.