5 Test measurements

A series of dedicated measurements was conducted with the GEM radiotelescope at 1465MHz during the present observational period in Brazil in order to improve the discrimination of the sky contaminating sources. The measurements consisted of pairs of observations taken at $Z=0\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ and at $Z=30\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ in sky directions away from the Galactic Plane (see Fig. 3). Each observation sampled the radiometric signal every 0.56 seconds over a few minutes while an approximate 15-minute interval elapsed between the $Z=0\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ and the $Z=30\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ samplings. In this manner, a total of 6 measurements were obtained over a nearly 3-month period. Although the absolute level of ground contamination in general will be somewhat different for different pairs, the mean difference between the two levels, $\bar\Delta T_{\rm A}$, can be used for comparison with the model predictions outlined in the preceding section.

   

Table 2: Transmission factor $Q_{\rm n}$

feed		Effective Shielding
regime		none	fence	halo	double
Fresnel		0.98	0.80	0.70	0.41
Fraunhofer		0.82	0.34	0.55	0.10

This differential measurement approach relies, however, on our ability to separate likewise the other constituents of the antenna noise temperature, namely, the atmospheric emission and the sky background. The latter is a mixture of synchrotron and free-free radiation originating in the Galaxy, Cosmic Microwave Background Radiation (CMBR) and a diffuse background of extragalactic origin. Depending on the sky direction Galactic emission at 1465MHz can be some 5 times larger or even a full order of magnitude smaller than the signal due to the CMBR. The atmospheric contribution, on the other hand, is necessarily larger at $Z=30\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ than at the zenith because of a larger air mass. At 1465MHz the bulk of the emission by the atmosphere is due to the pressure-broadened spectra of the O₂ molecule. Using the reference model proposed by Danese & Partridge (1989) (see also Liebe 1985 and Staggs et al. 1996) a straightforward secant law correction to the zenith contribution at the Brazilian site gives an estimate for the differential atmospheric component of $0.305\pm0.090$þK.

  
5.1 Data reduction

Our data was first time-ordered and corrected for thermal susceptabilities of the receiver baseline ( $0.3591\pm0.0007$ þK/ $\ifmmode^\circ\else\hbox{$^\circ$ }\fi$C) and fractional gain ( $0.00922\pm0.00001/\ifmmode^\circ\else\hbox{$^\circ$ }\fi$C). Then, 44.8 s bursts of 2.24 s firings of a thermally stable noise source diode were extracted from the data stream and used to calibrate the overall system gain. Table 3 summarizes the results of the observations along with the number of samples and the implied differences in antenna temperature between the two Z directions for: (i) the measurements, (ii) the Galactic emission background and (iii) the final budget (including the increase due to the larger optical depth of the atmosphere at $Z=30\ifmmode^\circ\else\hbox{$^\circ$ }\fi$). The Galactic contribution was estimated using a partial map (65.21 hours of data) of the sky signal from the GEM experiment at 1465MHz, whose baseline has been so far properly corrected according to a destriping algorithm in order to filter out low frequency noise (Tello 1997). The data for this map makes up about 30% of the data used in preparing the map in Fig. 3, but due to sampling differences (which bias the destriping process - see also Table 4) it has been split into the two maps shown in Figs. 15 and 16 along with the locations chosen for the paired measurements listed in Table 3.

Figure 15: Destriped partial map at $1.6\ifmmode^\circ\else\hbox{$^\circ$ }\fi$-pixel resolution of the high-sampled sky regions displayed in the declination band of Fig. 3. The antenna temperature range is 1.5þK in 60þmK contour steps. Squared and triangular symbols denote, respectively, the sky directions of the paired $Z=90\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ and $Z=60\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ observations for the test measurements

   

Table 3: Antenna temperature in the GEM experiment at 1465MHz for observations at $Z=0\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ and $30\ifmmode^\circ\else\hbox{$^\circ$ }\fi$

		$Z=0\ifmmode^\circ\else\hbox{$^\circ$ }\fi$			$Z=30\ifmmode^\circ\else\hbox{$^\circ$ }\fi$			$\Delta T_{\rm A}$þ(K)
pair		$T_{\rm A}\pm 1\sigma$þ(K)	N		$T_{\rm A}\pm 1\sigma$þ(K)	N		measurement	Galaxy	final budget
1		$10.798\pm0.042$	104		$12.151\pm0.035$	150		$1.353\pm0.055$	$-0.083\pm0.035$	$0.97\pm0.11$
2		$9.620\pm0.036$	146		$11.028\pm0.036$	93		$1.408\pm0.051$	$-0.254\pm0.066$	$0.85\pm0.12$
3		$11.212\pm0.040$	219		$12.180\pm0.046$	347		$0.967\pm0.061$	$+0.307\pm0.043$	$0.97\pm0.12$
4		$7.923\pm0.035$	71		$9.299\pm0.040$	148		$1.376\pm0.054$	$+0.173\pm0.050$	$1.24\pm0.12$
5		$8.922\pm0.030$	132		$10.151\pm0.031$	82		$1.229\pm0.043$	$-0.028\pm0.011$	$0.90\pm0.10$
6		$12.472\pm0.038$	131		$13.582\pm0.039$	267		$1.110\pm0.054$	$+0.185\pm0.024$	$0.99\pm0.11$

In order to extract the antenna temperature in a given direction, the pixel nearest to it was found first and then averaged with the surrounding set of 8 neighbouring pixels taken at half-weights. This procedure allows us to sample the sky in a square region $4.8\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ ( $1.6\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ per pixel) on the side and is consistent with a HPBW of $\approx 5.4\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ for the 1465MHz beam (Tello 1997). This can also be verified in Table 4 where we compare these estimates with those of the nearest pixel value itself and the average from the 4-pixel area enclosing the given direction along with the sampling differences among the different pairs. Note that pair 5 is actually missing in Fig. 15 and, therefore, we have provisionally supplemented the data in Tables 3 and 4 with the differential measurement obtained using the map in Fig. 3. To see that this is not as bad as it appears, the mean absolute difference between the estimates for pairs 1, 2 and 3 in the maps of Figs. 3 and 16 (low-sampled sky) is $0.237\pm0.066$þK, but only $0.112\pm0.040$þK for pairs 4 and 6 in the high-sampled regions of the map in Fig. 15. Thus, within the sensitivity of our measurements ( $\approx 20$þmK) the Galactic contributions to the differential measurements in Table 3 turn out to be smaller than, or as large as, the one estimated for the emission of the atmosphere.

   

Table 4: Effects of binning strategy for the Galactic contribution to differential measurements at 1465MHz ^a

		1-pixel				4-pixel matrix				9-pixel matrix
pair		$\Delta T_{\rm A}$þ(K)	N90	N60		$\Delta T_{\rm A}\pm 1\sigma$þ(K)	N90	N60		$\Delta T_{\rm A}\pm 1\sigma$þ(K)	N90	N60
1		-0.100	13	9		$-0.078\pm0.032$	48	39		$-0.083\pm0.035$	110	92
2		-0.288	4	63		$-0.257\pm0.026$	14	233		$-0.254\pm0.066$	33	338
3		+0.254	18	11		$+0.325\pm0.034$	74	56		$+0.307\pm0.043$	165	120
4		+0.222	70	47		$+0.217\pm0.023$	293	203		$+0.173\pm0.050$	697	428
5		-0.008	160	252		$-0.019\pm0.006$	637	997		$-0.028\pm0.011$	1419	2179
6		+0.178	63	85		$+0.177\pm0.023$	251	343		$+0.185\pm0.024$	568	784
$\bar\Delta T_{\rm A,\oplus}^{\rm obs}$		$1.003\pm0.071^{\mathrm{b}}$				$0.983\pm0.046\pm0.054$				$0.992\pm0.044\pm0.062$

^a

The entries referred to as N₉₀ and N₆₀correspond to the number of observations sampled in the determination of the sky signal observed toward $Z=90\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ and $Z=60^\circ$, respectively.

^b

Only the external error ( $\sigma/\sqrt{5}$) has been assigned in this case.

The weighted average of the values in the last column of Table 3 is an estimate of the differential ground contamination in the GEM experiment at 1465MHz. We obtain $\bar\Delta T_{\rm A,\oplus}^{\rm obs} = 0.992$þK with internal and external 1-$\sigma$ error estimates of 0.044 and 0.062þK, respectively (see also Table 4). Based on the ratio between these two errors, we can rule out the presence of systematic errors, which may have been introduced, for instance, by unaccounted stray radiation contamination of sidelobes other than those considered here. In fact, aside from the differential measurement approach, which reduces the effect of residual sidelobe contamination, the signal contrast of even the brightest sky features relative to that of the ground does not go above the 13-dB level. Only the presence of the Sun could offer potential problems, but except for pair 1, none of the other measurements was conducted with the Sun above the horizon. Still, the estimate from pair 1 does not raise suspicious concerns, even though the Sun was seen at $90.0\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ from axis and at $71.2\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ during the observations toward $Z=60\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ and $Z=90^\circ$, respectively.

  
5.2 Orientation of the $\phi$-plane

Before attempting a comparison of $\bar\Delta T_{\rm A,\oplus}^{\rm obs}$ with our model predictions, we need to assign the orientation of the $\phi$-plane of the feed in order to select the most likely profile. In addition, we have to apply the model calculations for the shield configuration actually used during the observations. Although the halo was the same as the one assumed to obtain the results in Figs. 10 through 13, the attenuation of the fence was increased by using a wire mesh with holes half as small and wires 25% thinner (according to our attenuation formula in Paper I we should thereby obtain a 6.2-dB screening effect at 1465MHz). Finally, the entire fence was raised 80þcm above the ground.

The orientation of the $\phi$-plane of the feed could be inferred by direct comparison of the feed diagram in Fig. 7b with the mapping of the beam pattern of the antenna by some convenient point source. This procedure is, of course, based on the assumption that the feed axis is also not perfectly aligned with the optical axis of the secondary for an asymmetric beam pattern to be projected onto the sky. In our case we chose the Sun, at a particular time of the year, which at the Brazilian site can be made to intercept the Galactic scans at $Z=30\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ with sufficient angular coverage ( $\sim 30^\circ$) around the beam axis. The result of such a mapping is displayed in Fig. 17 in 20 contour steps of 1þdB. The brightest region, corresponding to the precise passage of the scan circle through the Sun, could not be mapped up to a true 0-dB level because the signal overshot the detector threshold. This may have caused the double-lobed structure seen inside the main beam pattern in Fig. 7b to smooth out in the mapping of Fig. 17. In fact, in 1994, when the solar activity was relatively low, we recorded a solar transit (see Fig. 18) in Bishop, CA, which did not saturate the detector and did reveal a double-peaked main beam. In Fig. 17 the innermost contours follow the outlines of a bulged shape which is reminiscent of the double-lobed structure. Thus, together with the ellipticity of the surrounding contours in both diagrams we determined the $\phi$-plane orientation of the feed from the difference in the orientation of the major axis of these elliptical contours. The 10-dB contours are well confined inside elliptical boundaries with eccentricities of 0.64 and 0.34 for the feed and antenna patterns, respectively. The semi-major axis of the ellipse in the direction of the larger lobe in Fig. 7b is then oriented along $\phi=125\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ while that in the direction of the bulged region in Fig. 17 corresponds to $\phi^\ast=317^\circ$. Since $\phi^\ast\equiv\phi+180^\circ$, according to the system of coordinates used in Fig. 17, we obtain a $\phi$-plane orientation for the feed of $12^\circ$.

Figure 16: The low-sampled complement of the map in Fig. 15, but at the same resolution and with the same gray scaling in antenna temperature. The upper-right hand corner is data defficient due to $60\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ custom cuts around the Sun

Figure 17: Beam pattern mapping of the 1465MHz backfire-fed GEM antenna, in twenty 1-dB steps and at a pixel resolution of at $1.6^\circ$, using the passage of the Sun through the $Z=30\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ scan circles of the antenna on the 29-th of September 1999 in Cachoeira Paulista, Brazil. The $\phi ^\ast$ angle of the pattern is measured counterclockwise from the ordinate axis and corresponds to $\phi-180\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ in the coordinate system of Fig. 7b while the elevation of the Sun is given by $60^\circ-\theta\cos\phi^\ast$. The arrow indicates the major axis alignment of the 10-dB elliptical contour toward the larger component of the double-lobed structure in Fig. 7b

Figure 18: Antenna noise temperature record (solid line) of a solar transit in Bishop, CA, during data taking operations at 1465MHz in October 1994. The elevation of the Sun, $H_{\rm Sun}$, is given by the dotted line. The peaks A and B indicate relative maxima in the antenna response and span an $\approx 5\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ interval in the azimuth coordinate of the Sun

  
5.3 Final estimates

Our diffraction model predicts a differential ground contamination of $\Delta T_{\rm A,\oplus} = 1.380$þK for the shield configuration used during the observations and an orientation of $\phi_{\rm plane}=12^\circ$. In order to predict the observed value of $\bar\Delta T_{\rm A,\oplus}^{\rm obs}$, we have to adjust the attenuation coefficient of the wire mesh by an efficiency factor $\beta=0.675\pm0.052$ or, equivalently, increase the screening of the fence by 1.71^+0.35_-0.32þdB. The resultant profile has been included in Fig. 11. $\beta$ scales linearly not only with $\Delta T_{\rm A,\oplus}$, but also with the predicted differential ground contributions from the halo, $\Delta T_{\rm A,\oplus}^{\rm\,hal}$, and from the fence, $\Delta T_{\rm A,\oplus}^{\rm\,fen}$. So, if

 

$${\beta\over10^{-3}} = -155.880 + 837.363 \,\left({\Delta T_{\rm A,\oplus}\over {\rm K}}\right),$$ (1)

then the corresponding contributions from the halo and from the fence are

 

$$\left({\Delta T_{\rm A,\oplus}^{\rm\,hal}\over {\rm mK}}\right) = 31.49 - 169.25\,\left({\Delta T_{\rm A,\oplus}\over {\rm K}}\right)$$ (2)

and

 

$$\left({\Delta T_{\rm A,\oplus}^{\rm\,fen}\over {\rm mK}}\right) = 193.61 - 40.00\,\left({\Delta T_{\rm A,\oplus}\over {\rm K}}\right)$$ (3)

with 1-$\sigma$ errors of 0.03 and 0.01 in the zero-points and linear coefficient, respectively, in (2) and (3); but 1 order of magnitude smaller in those of (1).

These formulae tell us that, as the screening of the fence becomes less efficient ($\beta$ increasing), the differential ground contribution increases, even though the one from the diffracted components decreases. In this spillover-dominated scenario with $Q_{\rm n}=0.67\pm0.01$ (see Fig. 14) the ground contamination contributed by diffraction at the halo and at the fence will decrease with increasing Z as long as $\beta\,\raise 0.25ex\hbox{$>$ }\hskip -0.7em \raise -0.45ex \hbox{$\scriptstyle\sim$ }\hskip 0.2em\,0.00011$ and $\beta\,\raise 0.25ex\hbox{$>$ }\hskip -0.7em \raise -0.45ex \hbox{$\scriptstyle\sim$ }\hskip 0.2em\,3.9$, respectively. For most practical fences, the lower bound on $\beta$ implies that diffraction at the halo should always decrease with Z. In order to have the same scenario at the fence, the attenuation of the wire mesh would have to be quite low $(\,\raise 0.25ex\hbox{$<$ }\hskip -0.7em \raise -0.45ex \hbox{$\scriptstyle\sim$ }\hskip 0.2em\,0.3 \;{\rm dB})$.

Table 5 gives the refined model estimates of the ground contamination levels for GEM observations at 1465MHz in the Southern Hemisphere.

   

Table 5: Ground contamination in $Z=30\ifmmode^\circ\else\hbox{$^\circ$ }\fi$ GEM data at 1465MHz from the Southern Hemisphere

	sidelobe	shield a		contamination		error
				(mK)		(mK)
	spillover	double		975		75
	diffraction	fence		154		3
	diffraction	halo I		28		2
	diffraction	halo II		-11		1
	Total	double		1146		75

^a Estimates are given for a double-shielded scenario were the rim-halo contribution to the diffracted component has been separated into exposed (halo I) and hidden (halo II) portions as discussed in Sect. 4.2.