Annu. Rev. Astron. Astrophys. 1980. 18: 489-535 Copyright © 1980 by Annual Reviews. All rights reserved

(b) Antenna

The antenna design in background experiments is critical. The background radiation was discovered by Penzias & Wilson (1965) primarily because a very low side-lobe antenna had been constructed at Bell Laboratories for satellite communications. It is most unlikely that any general purpose radio antenna existing at that time could have been used to bake the discovery with any degree of confidence. The remaining observations in Table 1 were carried out with special purpose horn antennae.

Since the background radiation covers the entire sky, there is no optimum antenna beam size to investigate the spectrum unless one wants to avoid discrete local sources. Most of the spectrum experiments have used beam widths large by radio astronomy standards, 4° to 20°, but still small enough to allow zenith angle scanning to measure the atmospheric emission. The primary concerns are antenna thermal emission and the reduction of antenna diffraction at large angles to the optic axis - the side lobes. The importance of controlling the side lobes is shown by a simple calculation. Let the main beam, observing the CBR at 3 K, include a solid angle of 0.02 steradians (10° beam). The 300 K earth, airplane, or spacecraft fills the 2 steradian backplane. If the contribution from the backplane is to be less than 10% of the CBI, the point source response of the antenna for angles greater than 90° to the optic axis must be less than 10^-6 (-60 db) of that along the optic axis, and still smaller if the experiment is to measure the spectrum at high frequencies where the Rayleigh-Jeans approximation for the CBR is no longer valid. Since the off-axis response of the antenna is usually a steep function of the angle from the optic axis, the contribution from atmospheric emission, determined by scanning the zenith angle, can be confused by an increase in the ground radiation contribution from poorly suppressed side lobes.

Most of the antennas used in the low frequency experiments were single mode rectangular or cylindrical horns 10 to 15 wavelengths wide at their last diffracting surface. The field distribution at the horn mouth is close to cosinusoidal giving a far field angular response function at best varying as ~ (2n sin)^-4 at large angles where n is the aperture width in units of the wavelength and the angle to the optic axis. These horns were just barely good enough and most groups resorted to placing large metallic reflectors under and around the horn to reflect the "cold" sky into the antenna back lobes.

The high precision experiments pioneered by the Princeton group used a horn aimed at a mirror as the antenna. In this scheme the radiometer was held fixed throughout the experiment while the mirror was rotated to perform zenith scanning or removed for calibration. This technique eliminated a possible systematic error source arising from changes in the radiometer offset due to varying gravitational loading of the waveguide plumbing. However, the mirror is an additional polarization- and angle-sensitive emitting surface in the beam.

In the past few years substantial advances in low side-lobe horn designs have been made. The narrow-band corrugated horns used in the U2 experiment to measure the large angular scale anisotropy of the CBR (Gorenstein et al. 1978) have demonstrated a back-lobe rejection of 10^-8 (-80 db) (Janssen et al. 1979). Flared broad-band ultimode (apodizing) horns developed for balloon-borne spectrum measurements (Mather 1974, Woody 1975) have been designed with the Keller theory of geometric diffraction (Levy & Keller 1959, Mather 1980). The flared horns, tested at JPL, also exhibit back-lobe rejection of 10^-8 or smaller (Janssen & Weiss 1977).

The thermal emission by the antenna is not a serious problem at low frequencies and is measured by varying the temperature of the antenna while looking into a cold load or at the sky. In those experiments where the calibrator covered the entire antenna aperture, antenna emission does not enter the accounting procedure in first order. High frequency observations are more troubled by antenna emission both because the emissivity of the metal surfaces grows at least as fast as the square root of the frequency and because the CBR power is not rising as fast as a Rayleigh-Jeans spectrum. The best recourse here is to cool the antenna.