We present the design, manufacturing methods, and characterization of 20 microwave feed horns currently in use on the Microwave Anisotropy Probe (MAP) satellite. The nature of the cosmic microwave background (CMB) anisotropy requires a detailed understanding of the properties of every optical component of a microwave telescope. In particular, the properties of the feeds must be known so that the forward gain and sidelobe response of the telescope can be modeled and so that potential systematic effects may be computed. MAP requires low emissivity, azimuthally symmetric, low-sidelobe feeds in five microwave bands (K, Ka, Q, V, and W) that fit within a constrained geometry. The beam pattern of each feed is modeled and compared with measurements; the agreement is generally excellent to the -60 dB level (80 degrees from the beam peak). This agreement verifies the beam-predicting software and the manufacturing process. The feeds also affect the properties and modeling of the microwave receivers. To this end, we show that the reflection from the feeds is less than -25 dB over most of each band and that their emissivity is acceptable. The feeds meet their multiple requirements.

We describe two related experiments that measured the anisotropy in the cosmic microwave background (CMB). QMAP was a balloon-borne telescope that flew twice in 1996, collecting data on degree angular scales with an array of six high electron mobility transistor-based amplifiers (HEMTs). QMAP was the first experiment to use an interlocking scan strategy to directly produce high signal-to-noise CMB maps. The QMAP gondola was then refit for ground based work as the MAT/TOCO experiment. Observations were made from 5200 m on Cerro Toco in Northern Chile in 1997 and 1998 using time-domain beam synthesis. MAT/TOCO was the first experiment to see both the rise and fall of the CMB angular spectrum, thereby localizing the position of the first peak to l_{peak}=216 +/- 14. In addition to describing the instruments, we discuss the data selection methods, checks for systematic errors, and we compare the MAT/TOCO results to those from recent experiments. We also correct the data to account for an updated calibration and a small contribution from foreground emission. We find the amplitude of the first peak for l between 160 and 240 to be T_{peak}=80.9 +/- 3.4 +/- 5.1 uK, where the first error is statistical and the second is from calibration.