The BIMA receivers ( drawing
) use a straight through optical path.
The dewars are designed to accomodate 4 frequency bands, but
at present only two are implemented - a 3mm band (70-116 GHz)
and a 1mm band (210-270 GHz).
The receivers are sensitive to a single linear polarization (vertical).
Polarization plates fabricated from Rexolite (a cross-linked
polystyrene) are mounted on a motorized frame in front of the receiver.
The array control computer can switch them in or out of the beam
in ~ 3 seconds. The polarization plates are mounted in pairs - one
converts from right circular to linear, the other from left circular to
linear. In addition, the frame provides an empty slot for normal
observing - no plate in the beam.
If the plates are mounted exactly normal to the beam axis, reflections
cause large standing waves in the crosspolar response. We mount
the plates at about an 8 degee tilt on the frame to reduce this unwanted
The plates are fabricated using a slitting saw on a computer-controlled
milling machine. As shown in the drawing (113 GHz quarter wave plates ),
slots are made in both sides of the plate to help reduce reflections.
The measured losses are higher than we expected, and suggest either
that the Rexolite has a loss tangent of ~0.001, or perhaps that scattering
from the slots (approx 3/wavelength) is significant.
At present BIMA has 3 sets of quarter wave plates, centered at frequencies
of 89 GHz, 113 GHz, and 230 GHz. There are also half wave plates (deeper
grooves, +- 22.5 degrees from vertical) to flip between two orthogonal
linear polarizations (+- 45 degrees from vertical), which were built
for Zeeman measurements of CN at 113 GHz. The crosspolarization,
or leakage terms, are quite stable with time. BIMA memo 89 summarizes
measurements of the polarization leakage terms for the 1mm wave
plates over a 2 year period.
To make measurements of all 4 Stokes parameters, one moves the
plates on different antennas in a walsh function switching pattern
to obtain measurements of all 4 cross-polarizations (RR, RL, LR, LL)
on each baseline as rapidly as possible. Further details are given
in Rao et al., 1998 (ApJ, 502, L75).