Submillimeter Array

Did You Know?

• The SMA project began in 1983 as part of an initiative by Irwin Shapiro, then new director of the SAO.
• The SMA was used to discover that Pluto is 10 K cooler than expected and was the first radio telescope to resolve Pluto and Charon as separate objects.
• The SMA is part of the Event Horizon Telescope, which produced the first image of a black hole.

Array Design

The SMA was built northwest of the saddle between the cinder cones Pu'u Poli'ahu and Pu'u Hauoki, about 140 meters below the summit of Mauna Kea. In 1996 Eric Keto studied antenna placement for the array and found that the most uniform sampling of spatial frequencies, producing the cleanest point spread function, was achieved when antennas were arranged in the shape of a Reuleaux triangle. Consequently, pads for the antennas were laid out to form four Reuleaux triangles, with the easternmost pad serving as a shared corner for all four. However, the lava field site contained many rocky ridges and depressions, preventing exact placement at the optimal positions. Typically all eight antennas are deployed on pads forming one Reuleaux triangle, yielding four configurations named subcompact, compact, extended, and very extended. Antenna moves follow a roughly quarterly schedule determined by approved observing proposals. A custom transporter lifts an antenna off a pad, drives it along dirt roads, and places it on a new pad while maintaining power to the cryogenic receivers. Each pad has a conduit to the central building carrying AC power cables and optical fibers. Multi-mode fibers handle low bandwidth digital signals, while Sumitomo LTCD single-mode fibers, with extremely low thermal expansion at typical subsurface temperatures on Mauna Kea, carry reference signals for the local oscillator and return the IF signal, allowing operation without closed-loop delay measurements.

Antennas

Each of the eight antennas has a 6 meter primary mirror made of 72 machined cast aluminum panels, chosen over carbon fiber due to concerns about snow and volcanic dust damage. The panels, each about 1 meter wide, were machined to 6 microns accuracy and are supported by a carbon fiber tube backup structure enclosed by aluminum panels. Panel positions can be adjusted from the front of the dish. Initial surface adjustment was done in the service hangar using a rotating template; after deployment, near-field holography with a 232.4 GHz beacon on the Subaru building guided further adjustments, repeated periodically. After several rounds, the surface error is typically about 15 microns RMS. Heating units on the primary mirror, secondary mirror support, and secondary mirror itself prevent ice formation in high humidity. Each antenna has a temperature-controlled cabin housing control electronics and Nasmyth focus receivers, nearly enclosing the steel mount to minimize thermal pointing errors.

Receivers

The SMA uses cryogenic SIS heterodyne receivers at a bent Nasmyth focus, all mounted in a single large cryostat within the antenna cabin. The cryostat can hold up to eight receiver inserts, each containing one receiver. A rotating wire grid beam splitter and rotating mirror direct the two linear polarizations to two receiver inserts, allowing observation of either a single polarization of two different frequency bands simultaneously, or both polarizations of a single band for improved sensitivity and Stokes parameter measurement. Receivers cover 194 to 408 GHz without gaps, but full polarization measurements are possible only around 230 and 345 GHz, where pairs of receivers can be tuned to the same frequency and quarter wave plates inserted. The receivers are sensitive to both sidebands; sidebands are separated by introducing a Walsh pattern of 90 degree phase changes in the LO signal and demodulating it in the correlator. A Walsh pattern of 180 degree phase changes, unique to each antenna, suppresses crosstalk between IFs. With the recent wideband update, two receivers tuned 12 GHz apart allow observation of a 44 GHz wide interval of sky frequencies without gaps.

Correlator

The original SMA correlator correlated 2 GHz of IF bandwidth per sideband from two active receivers in eight antennas, producing spectral data for 28 baselines. Analog-to-digital converters sampled at 208 MHz, so the IF was downconverted into 24 partially overlapping 104 MHz wide chunks before sampling. Data were sent to 90 large PC boards, each holding 32 ASIC correlator chips in an XF design; in the default configuration 6144 lags were calculated for each of two receivers on 28 baselines, then an FFT converted lag data to spectra with 812.5 kHz per channel resolution. The correlator could be reconfigured for higher resolution on certain chunks. The chips were designed at MIT Haystack and funded by five institutions. The correlator could also correlate all 45 baselines when including the CSO and JCMT, but for only one receiver per antenna. In 2016 a new correlator called SWARM came online, increasing total IF bandwidth and sensitivity. SWARM is an FX design using 4.576 GHz ADCs and Xilinx Virtex-6 SX475T FPGAs on ROACH2 boards from CASPER. It operates at a single spectral configuration of uniform 140 kHz per channel resolution across the entire bandwidth, and data are stored at this high resolution even for low-resolution projects. Each quadrant processes 2 GHz of IF bandwidth per sideband for two receivers in all eight antennas; when receivers are tuned to the same frequency, full Stokes polarization is calculated. The full correlator has six SWARM quadrants, allowing 12 GHz of bandwidth per sideband for two receivers on all baselines, yielding 48 GHz total sky frequency coverage.