As SETI researchers are quick and keen to point out, the Allen Telescope Array, currently under construction about 200 miles northeast of San Francisco, is the first professional radio telescope designed from the get-go to speedily search for extraterrestrial signals. When completed, it will comprise 350 antennas, spread over roughly 150 acres of lava-riven real estate.
That's a lot of antennas and a good chunk of property, too. But exactly where on that acreage will the antennas be erected? In other words, how will the array be arrayed? It will come as no surprise when I tell you that there's a siting plan: a topographic map covered with 350 dots indicating where each antenna will be located. But whenever I unroll the plan under the nose of an interested party, a thought balloon with a question mark inflates above their head. "That looks so random ..." they blurt out.
And indeed it does. The plan resembles the holes on a dartboard – and one used by a less-than expert player at that. Folks are dismayed because they assume that the antennas should be on some sort of regular grid.
The reason for the (apparent) randomness is this: since the Allen Telescope Array will be used to map nebulae and galaxies in addition to its tasks as a SETI search machine, it should be a good radio "camera". In other words, it should be able to make detailed, and high quality, radio maps of the sky.
To do that, you really need a giant antenna dish that's a half-mile or more in size. That's not only expensive, it's just about impossible to build.
But in the 1960s, radio astronomers in Cambridge, England figured out that you could simulate the behavior of a behemoth antenna with smaller ones if you moved them around, collecting data at various antenna positions. It was OK to spend weeks doing this, since the object being "photographed" – a galaxy, for instance – doesn't move or change noticeably during a multi-week exposure. This technique, appropriately (if formidably) called "aperture synthesis", is nothing more than a scheme for breaking an impractically large antenna into smaller parts, collecting data piecemeal, and then combining the information with the help of a computer to form a picture.
Fair enough. But if you wish to make a high-quality image, you need to measure the incoming signal at as many different antenna spacings as possible. This means garnering data from antennas that are practically shoulder to shoulder, as well as from antenna pairs that are a half-mile or more apart. If you site the antennas on a regular grid, you'll have too many pairs with the same separation, a decidedly inefficient approach.
Getting the Point
Astronomers evaluate their siting plans by measuring the array's so-called "point spread function." This may sound like a technique for winning at blackjack, but in fact it's a scheme routinely used for testing optical systems such as telescopes and camera lenses. The point spread function (PSF) is nothing more than the picture the instrument would make of a point source – a radio transmitter (or a light bulb, in the case of optical instruments) that's very, very far away. In a perfect world, this would be, well, a point. In the real world, it will be at least a little blobby, and might also have rings and things around it (for those jazzed by jargon, these are referred to by radio engineers as sidelobes). If your PSF's really second-rate, the pictures from your instrument will resemble a Seurat painting: not much detail.
Douglas Bock, who until recently was the System Scientist for the Allen Telescope Array, did the heavy lifting for designing the ATA's siting plan. He knew he had 350 antennas to plant, and his brief was to make the cleanest, sharpest PSF possible. Of course, there were a few constraints. "Obviously, the antennas had to fit on the property we could rent from the land owner, as well as on land held under a special use permit from the Forest Service. And that's not a nice round or square shape but a sort of funny one," he says.
"In addition, there were some practical considerations. I had to avoid the edges of lava flows, and skirt around existing buildings, roads, and trees."
But within these constraints, Bock could proceed as follows. He began with a random placement of antennas and then, using appropriate software, computed an indicator of how good the resulting PSF would be. "I then looked at each antenna in turn, and decided whether moving that dish would make the resulting image better or worse. If better, I moved it."
Much computer time later, Bock had a point spread function that was tack sharp, and free of noxious sidelobes. He had designed an instrument that, in a matter of minutes, could make high-quality images of cosmic radio sources virtually any place on the visible sky.
This is performance that outclasses just about any other radio telescope. Why? Well, you can chalk it up to one obvious, but often overlooked, feature of the Allen Telescope Array. It has lots and lots of antennas. New Mexico's Very Large Array, an instrument often featured in movies and TV shows, has 27 antennas, located along three, equally-spaced splines (like flattened tripod legs). It can make great radio pictures, but it takes eight hours of observing to achieve top-notch quality, because it relies on the rotation of the Earth to arrange its antennas in enough "spacings" to generate a clean image. The Westerbork Synthesis Radio Telescope, in Holland, has 14 antennas in a straight, east-west line. It requires 12 hours of Earth's spin to make a good picture of the sky.
So when you see renditions of the Allen Telescope Array, with its seemingly randomly spaced antennas looking as if the dishes were dropped from a high-flying cargo plane, keep in mind that this apparent lack of order is misleading. Just as for the trading floor of the New York Stock Exchange, a chaotic appearance belies what is a singular, and finely tuned system.