(large size)

When optical scientist Rick Lyon worked on a computer technique to align the Hubble Space Telescope, a backup for the on-board optical control sensors, he never thought NASA would need to use it. Tragically, the original, blurred Hubble images of 1990 portended a problem beyond the sensors' operating range.

Using high-performance computing, Lyon, who worked for Perkin-Elmer Corp. at the time, and NASA Goddard Space Flight Center (GSFC) researchers enhanced the technique enough to determine the optical prescription to correct Hubble. Joining together at GSFC in 1994, the team further increased the fidelity and completely removed the flawed telescope's effects in observed images.

Today, with support from NASA's High-Performance Computing and Communications (HPCC) Program, they are applying these same methods to future NASA space flight missions, among them the Next Generation Space Telescope.

Correcting near-sightedness

The original Hubble flaw stemmed from the manufacturer's grinding the primary mirror "slightly too flat," said Lyon, who is affiliated with GSFC through Universities Space Research Association's Center of Excellence in Space Data and Information Sciences (CESDIS). The mirror being off by .002 millimeter at the edge made the telescope "near-sighted, "so distant points of light looked extended and fuzzy.

In one such effort, the GSFC team deblurred and restored an oversaturated, blob-like image of R-Aquarii, a system containing two orbiting stars and a gas jet. Led by astrophysicist Mike Hollis, they started with an observed point-spread function (PSF). A PSF is how the telescope responds to a point source of light; a superposition of many PSFs makes up an observation.

Computer scientist John Dorband then implemented a mathematical algorithm known as Maximum Entropy to extract the observed PSF from the R-Aquarii image. He applied the 16,384-processor MasPar-2/1 high-performance computer to the 250,000 data elements. The results: "We were able to reconstruct regions that were overexposed, as well as correct for the flawed Hubble primary mirror," Dorband said.

Using an observed PSF is less than ideal because it contains random effects. Only a computed PSF based on a perfect optical model is fully reliable. To do just that, Lyon, Hollis and Dorband formed the multidisciplinary Optical Systems Characterization and Analysis Research Project, or OSCAR for short. NASA HPCC's Earth and Space Sciences Project supplies OSCAR's core funding.

Perfecting the Hubble model was a natural job for phase retrieval. "You build a computer model of every optical element, every mirror and every surface the light hits going through the telescope," Lyon said. "You then simulate the light propagating through the telescope and adjust the computer model until the simulated images match the observed images to determine what is wrong with the telescope."

Building the next generation

Trash no more, OSCAR-restored images later agreed with those observed by Hubble's corrected optics. Success in characterizing other telescopes showed that "ours is not a one-instrument system but a technique potentially applicable to many NASA missions," Dorband said. Their work caught the attention of Next Generation Space Telescope (NGST) planners given the task of building a telescope four times Hubble's size for one-quarter its $2 billion cost.

These OSCAR images simulate computationally supporting the coronagraph, a plug-in camera proposed for the Next Generation Space Telescope (NGST). In a solar system 80 light years away, light from a bright star (left panel) is extracted to reveal a planet (arrow on right panel).

NGST is part of NASA's Origins Program, which is aimed at discovering "how we got here," said GSFC's John Mather, NGST project scientist. Planned for launch in 2007 or 2008, the telescope will orbit the sun 1 million miles from Earth. NGST's location and 25-foot mirror will allow it to peer far enough to view the universe within a few hundred million years of the Big Bang. Infrared capabilities will penetrate dust clouds observed by Hubble and see stars being born. Mather expects NGST to start unraveling "mysteries as to whether galaxies or stars formed first."

The mirror "will not stay the right shape" under those conditions, Mather said. To keep things in line, NGST's basic design calls for a segmented, flower-shaped primary mirror; tens of actuators will push, tip and tilt the segments. The mirror will be imaged onto an active optic with hundreds of actuators to further compensate for mirror distortions. To faithfully capture images in this manner, NGST optical control software must perform "nearly one trillion floating-point operations," Lyon said.

OSCAR has adapted its phase retrieval software to "give NGST one NASA in-house version of the technology to do precise optical control," said CESDIS' Tim Murphy, a new OSCAR recruit. The project is also investigating whether more science can be accomplished using an on-board high-performance computer. This fall, the software will run on a testbed computer being built by NASA HPCC's Remote Exploration and Experimentation (REE) Project [see "Losing limits in space exploration," INSIGHTS, November 1998].

The testbed's fault-injection capabilities will simulate the cosmic rays of space, crashing the computer many times a day to see how well software equipped with REE-developed fault tolerance can restart, then identify and correct errors. Fault-tolerant software running on board "would allow you to have better images in real-time right from the telescope," Dorband said.

The optical concept itself will soon be explored in NGST's Deployable Cryogenic Active Telescope Testbed, a proving ground for hardware and several versions of sensing and adjustment methods from NASA's Jet Propulsion Laboratory (JPL), GSFC and industry. OSCAR is funded to compare methods of gathering information about the mirror distortions.

"When light goes through the telescope, it is like a rock dropping into water," Lyon explained. "You go far enough out, and the waves look like perfect circles, but there are distortions introduced by the telescope. We want to determine the shape of the telescope-distorted light waves." To gain broader understanding, Murphy has begun studying how different imaging configurations affect light in OSCAR's optics laboratory.

Besides the primary mirror, NGST will have plug-in instruments for specialized research. JPL leads a study on a coronagraph, a camera that would tune the mirror to look for planets in other solar systems. OSCAR is simulating this process using GSFC's 128-processor HIVE computer [see "Beowulf lives on -- as a build-it-yourself computer," INSIGHTS, November 1998].

More information is available at the following URLs: http://jansky.gsfc.nasa.gov/OSCAR/ and http://www.ngst.nasa.gov/