Pitch, roll, focus correction at multiple frequency on 3C48, CTA21
The main reflector surface (along the main
support cables) was measured and the panels adjusted in jan01 (see reflector
adjustments jan01). The gain improvement is shown in: sband
The remaining losses are from the unadjusted panels and the pitch, roll,
and focus collimation errors. Doing multi frequency measurements you can
estimate the surface rms using the ruze formula. This assumes that the
dish errors are not correlated over large areas, and that the other losses
(za spill over, collimation errors) have been corrected.
Using a theodolite and tilt sensors we've created
a model of the pitch, roll, and focus errors of the dish. We can
use the tiedowns to compensate for these errors (over a limited range)
while we track a source. This correction is currently limited by:
The pitch, roll, and focus is not measured very well above 16 degrees
The pitch and roll errors above 10 degrees za require a tiedown throw larger
than available. This requires us to try and reshim the elevation rails
On 18feb01 we tracked 3C48 (J0137+331) and CTA21 (J0318+164)
with pitch, roll, focus (PRF) enabled. The standard 4 strip calibration
technique was used. We alternated between lband, sband, and cband every
pattern. We wanted to find out :
When interpreting the data it is important to keep in mind sources of possible
error. These include:
What is the gain with PRF correction enabled.
Can we estimate the surface rms (all panels) assuming the PRF correction
is done correctly.
How accurate is the PRF model that we are using.
show the results of the measurements:
The correctness of the cal values affects the gain and ruze formula computation.
The 5400 Mhz cal value is probably in error.
The PRF model uses the temperature as one of the inputs to determine the
platform height. Experimenting before the run showed that the maximum
gain may have been occurring about .5 inches lower (for cband) than what
The fluxes need to be known at the different frequencies. Errors in them
will be errors in the gain.
Figure 1 shows the gain versus zenith angle for all of the measurements.
3C48 is a + while CTA21 is a *. The color encodes the different frequencies.
The solid line is source rising and the dotted line is source setting.
Since 3C48 is at 33 degrees declination it never came lower than 15 degrees
za. The first point of each source (solid line rightmost point) was done
with the PRF correction turned off. You can see the improvement the PRF
correction gives. For CTA21 the first point is a little lower, while the
first point for 3C48 is much lower at all frequencies. There is no zenith
angle gain correction included here. It should start at 15 degrees za and
drop off about 1 db at 20 degrees za. So the lband improvement 1st -> 2nd
point is probably the za change. The sband, cband improvement is much more.
Figure 2 plots the gain versus frequency for all the points (the first
point with no PRF correction has been left out). The lband gains for 3C48
and CTA21 vary by about 10%. The cband gains for the two sources also have
an offset approaching 10% while the sband gains for the two sources are
very close. This variation is not inaccuracies in the cal values since
we used the same cal value for each source. The discrepancy could be the
zenith angle difference or an error in the spectral index of the source
The solid lines are the computed gains for a system
with 10.5 K/Jy and a surface rms of 3 mm, 4 mm, 5 mm, and 6 mm (using the
ruze formula). The data points don't line up very well with any of the
curves (plus the constant should probably have been 12 K/Jy for a perfect
dish). This discrepancy could come from:
If you look at CTA21 a 4 mm surface rms is reasonable guess of the surface
the cal values used are not correct at the various frequencies
The fluxes are off
We aren't doing the PRF correction correctly or there is a larger zenith
angle gain dependence than we think.
The unmeasured panels have some large scale correlation's in their errors.
The different horn illumination patterns is affecting the gain.
Figure 3 shows the improvement in gain at cband for the PRF correction
on 3C48. The solid black line was measured on 02feb01 after the surface
was adjusted and with no PRF correction. The red line is the gain measured
on 18feb01 with the PRF correction enabled. The rightmost red point had
no PRF correction so it should have fallen on the black line (it's pretty
close). The green line shows what the PRF model predicts if the 02feb01
data was corrected for PRF errors. The two measured points
about transit (red) are close to the predicted (green) values. The predicted
points (green) below 160 deg az and above 210 deg az are not realistic.
The lower graph has the pitch and roll errors (.1
degree units) and the focus error in inches. The dashed line is the correction
factor that is being applied for the pitch and roll combined (blue dash
line) and the focus error (red dash line). The large focus
error correction is causing the gain values below 160 az to shoot up. We
probably need to have a better estimate of the pitch, roll, and focus errors
before we contemplate reshimming the elevation rails at high zenith angle.