The Hubble Space Telescope, launched into orbit 25 years ago, changed our perception of the universe.
The European-Extremely Large Telescope, or E-ELT, will capture images 15-times sharper than the dazzling shots Hubble regularly beams to Earth thanks, in part, to GPUs.
Almost half the length of a soccer field and with a 100-meter-tall dome on a rotating structure weighing 2,800 tons, the E-ELT “compares in size with the pyramids in Egypt,” said Damien Gratadour, associate professor at Universite Paris Diderot and LESIA, Observatoire de Paris, speaking earlier this year at NVIDIA’s GPU Technology Conference.
The E-ELT is expected to be completed atop Cerro Armazones in Chile’s Atacama Desert in 2024. Its 40-meter (131-foot) diameter telescope will dwarf the Hubble’s, which measures 2.4 meters (7.9 feet). But one of the biggest headaches for astronomers using Earth-based telescopes is seeing through atmospheric turbulence.
Our atmosphere is a mixture of air at different temperatures in perpetual movement, each with their own refractive index. This distorts what astronomers call the wave-front — the available light from stars — which reduces image quality, and thus the sensitivity of a telescope.
So no matter how big a telescope is, image resolution will be limited by the atmosphere. Perfectly transparent and stable skies are an impossibility, so astronomers find ways to compensate. And this is where GPUs can help.
Gratadour is part of a team deploying GPUs to perform close to real-time simulations of “multi-object adaptive optics” at an unprecedented scale. The technique improves image quality and compensates for the changes in a wave-front’s shape.
The team’s goal is to develop a large instrument, called MOSAIC, to equip the E-ELT and provide astronomers the biggest eye on the universe ever built.
MOSAIC will use adaptive optics to “reshape” the wave-front as it comes from a star or galaxy, crosses the atmosphere, and distorts. In a highly compute-intensive process, astronomers will measure the volume of turbulence and use a tomographic reconstructor to adapt the shape of deformable mirrors in the telescope to compensate for the effects.
MOSAIC’s performance was recently simulated with a system that includes four NVIDIA Tesla K20c GPUs. The result: the most realistic snapshot ever made of observations of distant galaxies with the E-ELT.
With this success under their belt, the team is now considering using GPUs to drive the instrument in real time when it will be in operation on the E-ELT — and potentially give astronomers the best view they’ve ever had of the cosmos.
For a more detailed explanation of Gratadour’s work, watch his presentation at GTC.