LSST: A New Telescope Concept
Moore's Law
In 1965, Gordon Moore, co-founder of Intel,
noted that the number of transistors per
square inch on integrated circuits had
doubled every year since the integrated
circuit was invented. He predicted that
this trend would continue well into the
future. To date, the density of circuits
has doubled about every 18 months, and most
experts expect this to continue for at
least another two decades. Since integrated
circuits are the basis of both computer
memory and processors, Moore's Law is
roughly equivalent to saying that the
capability of computers will double every
18 months for the forseeable future.
Remarkably, disk storage technology has
outpaced even Moore's Law over the past
decade.
The LSST Camera
In order to take advantage of high-quality
images produced over such a wide field, the
LSST's camera will contain over three
billion pixels of solid state detectors.
For the past twenty years astronomers have
employed digital image sensors known as
Charge Coupled Devices (CCDs) to great
effect. These devices can be made more than
90% efficient in detecting light (about 100
times more efficient that photographic
film) and at the same time designed to
introduce very little extraneous noise into
the detected signals. They may well be the
best choice for LSST.
Shown above is a design for LSST's
camera, including optical windows and
filters. The cutout shows the
camera's inner dewar (refrigerated
chamber) with its cooled focal plane in
place. Advances in microelectronics permit
low-power onboard electronics for each of
the 189 imager modules. Massively-parallel
read-out of these modules will generate up
to 20 terabytes of data per night.
GIGA-CAMERA: A NEW DISCOVERY SPACE
Astronomers, like oceanographers, have stumbled onto their discoveries by taking samples. We look at the sky with blinders: narrow fields of view which often require many hours of exposure. Occasionally we find something new. With luck we later follow up these chance single discoveries on other telescopes, if the objects are still emitting light. We are thus guaranteed to overlook a vast area of exploration: all objects which are faint, rare, change their brightness, or move. This is an enormous discovery space to miss. LSST will change all this. It will have the unique capability of imaging the entire visible sky to unprecedented faint limits multiple times per month. In a radical shift in paradigm, LSST will follow up its own discoveries.
The proposed explorations which will drive the development of the novel LSST facility have one requirement in common: the need to image wide swaths of the sky faint and fast. In optics there is a figure of merit for this capability, called "throughput" or "etendue", the product of the telescope capture area in square meters and the camera field of view in square degrees. Previous attempts to maximize throughput have focused on one or the other of these quantities.
Building telescopes of huge aperture has resulted in great light-gathering power, but at the expense of limited fields of view. Smaller telescopes with very wide fields have been constructed, but with limited apertures. These limitations are imposed by optics. The requirement of crisp images makes it impossible in traditional optical systems to achieve large throughput by simultaneously having large aperture and large field of view. LSST breaks this logjam. With its novel three-mirror optics and gigapixel camera, LSST will have an throughput of 319 meter^2 deg^2. This represents a fifty-fold increase over the best wide-field capability currently available, and makes possible the novel astronomy LSST will pursue. LSST is thus far more than a telescope. At the core of LSST is a camera with over three billion pixels. Driven by the need to capture faint, crisp images over the entire ten-square-degree field of one exposure, this will be the world's largest imager. While feasible by present technology, no imager this large has ever been attempted.
THE ORIGIN OF TODAY'S pervasive electronic imager from the familiar digital snapshot camera to the hundred-megapixel, wide-field imagers used in astronomy was a device invented in 1970 for the purpose of storing an audio message. Within hours of hearing of the need for a solid-state scrolling memory, George Smith and Willard Boyle at Bell Labs invented the charge-coupled device (CCD) using silicon integrated-circuit technology: a "bucket brigade" for electrons.
While fifty times more efficient at detecting light, the first CCDs covered 4000 times less area than previous detectors: photographic plates. Only in the last decade have electronic imagers, in the form of mosaics of many CCDs, grown large enough to be useful for LSST prototype wide-area surveys. The development in the 1990s of the four-to eight-megapixel CCD used in these mosaics depended on decades of R&D and improvements in microelectronics. This R&D has continued, enabling new types of sensitive electronic detectors. In addition to a new generation of panchromatic-sensitive CCDs, we now have CMOS (Complimentary Metal-Oxide Semiconductor) arrays of equal sensitivity. These low-noise, self-shuttering CMOS devices clock the photo-electrons down through several transistors under each pixel, rather than the bucket brigade use in CCDs.
Existing wide-field telescopes and cameras are stalled at a two- to four-meter telescope aperture and a fraction of a square degree per exposure. There are several reasons for this. Larger fields of view have proven impossible using the traditional one- or two-mirror plus multilens-corrector telescope optics. Hundred-megapixel CCD mosaics using early 1990's technology have hit a size limit imposed by wiring and electronics complexity and heat dissipation.
The LSST camera will use a different approach, similar to state-of-the-art microelectronics. All the control and processing electronics will be co-located with the individual detectors, thus avoiding a wiring nightmare. These hybrid building blocks will employ either the new generation CCD or CMOS photodetector arrays hybridized to an underlying CMOS ASIC (Application Specific Integrated Circuit). These megapixel imager modules will dissipate very litte heat into the cooled camera body and will be easy to replace if they fail. Up to one thousand of these modules populating the 64-centimeter diameter field of view will supply a parallel data stream. LSST's camera will produce 20 million megabytes of data every night. By 2006, Moore's Law (see sidebar) ensures that data processing and analysis hardware will routinely handle this data rate. Equally exciting is the progress made by recent surveys in automatic data-pipeline software. Breakthroughs in large optics, microelectronics, and software have come together. A new view of our universe will be the result.