HIGH-SPEED IMAGERS: ‘Lucky imaging’ is no accident

One of the most important developments in ground-based astronomy in recent years has been the development of adaptive optics, which has raised the resolution of terrestrial telescopes to near that of the Hubble Space Telescope at a fraction of its cost.

Oct 1st, 2006
Th 0610lfw Ne02

One of the most important developments in ground-based astronomy in recent years has been the development of adaptive optics, which has raised the resolution of terrestrial telescopes to near that of the Hubble Space Telescope at a fraction of its cost. Now, researchers at the Max Planck Institute for Astronomy (MPIA; Heidelberg, Germany) have gone one more step toward reducing the cost of near-diffraction-limited astronomical telescopes by developing a simple camera-and-software package that is potentially usable on any large high-quality ground-based telescope.

While the Hubble telescope solves the problem of atmospheric turbulence simply by spending its working life in the vacuum of space, adaptive optics (when paired with an earthbound telescope) correct for turbulence with a complex combination of wavefront sensors, deformable mirrors, secondary optics, software, and sometimes laser-created artificial guide stars. The expense of a complete adaptive-optics system for a large ground-based telescope-for example, the Keck Telescope-is low only in comparison to the billions spent on the Hubble.

‘Lucky’ choices

The MPIA researchers forgo the complex hardware of adaptive optics, instead relying on a high-sensitivity image sensor. They use a technique they call “lucky imaging,” in which the useful images are snapped during the brief stretches of time when the image is least affected by turbulence. The high sensitivity of the camera compensates for the fact that they are using only a small portion of the total incoming light (typically about 1% of the total data enters the final selection).

The researchers have installed a read-noise-free camera and a filter wheel-the combination of which they call the AstraLux-onto an existing 2.2 m telescope at the Calar Alto observatory, located in the Sierra de Los Filabres north of Almeria, Spain. “To attach the AstraLux camera to the Cassegrain focus of the telescope, we used an existing adapter,” says Stephan Hippler, one of the researchers. “The only change we had to make was implementing a Barlow lens to get the correct plate scale for AstraLux (the plate scale was chosen to Nyquist-sample the diffraction-limited point-spread function of the telescope).”

The camera, a front-illuminated EMCCD (electron-multiplying CCD) made by Andor Technology (Belfast, Ireland), detects single photons without using an intensifier. The camera is capable of taking 700 frames per second and has a peak quantum efficiency of 45% at 700 nm, near the 900 nm wavelength most often used by AstraLux. A combination of fast readouts and very low noise is what makes the technique of “lucky imaging” possible, note the researchers.

“An easy way to select ‘lucky images’ is to calculate the Strehl number of the reference star in each image taken,” says Hippler. “Depending on the requested final image angular resolution (this depends on the astrophysical program and research question), a minimum Strehl number of, let’s say, 0.1 is required. The selection algorithm then drops all images with Strehl numbers below 0.1. Another method is to calculate the Strehl number of all images and then select the 1% (or 5%, or 10%) best (in terms of Strehl) images. The images meeting the selection criterion are then ‘drizzle combined’ to reconstruct the diffraction-limited image.” (Drizzle combination, more formally known as variable-pixel linear reconstruction, can remove the effect of geometrical distortions in individual contributing images.1)

The angular field of AstraLux is 24 × 24 arcsec, considerably smaller than the Hubble’s 202 × 202 arcsec field-a function of the imager’s smaller pixel count (512 × 512) in comparison to that of the Hubble’s camera (4096 × 4096). But a wealth of astronomical objects are still accessible to AstraLux, including binary stars, planetary nebulae, young stellar objects, and globular-cluster centers. The 900 nm effective wavelength of AstraLux complements the typical 2.2 µm wavelength of adaptive-optics systems, notes Hippler.

First light for AstraLux was achieved in July of this year, a mere six months after the idea was conceived-a testament to the simplicity of the hardware. In one of the imaging system’s first uses, astronomers looked at the globular cluster M15, a relic from the early years of the Milky Way galaxy (see figure).

A conventional telescope image of the globular cluster M15 (5000 frames, 60 ms exposure time for each) at the Calar Alto observatory shows a point-spread function with a full-width at half-maximum (FWHM) angular size of 730 × 430 milliarcsec, considered very good at Calar Alto (top). A tip-tilt-corrected image of all 5000 frames has a FWHM of 390 milliarcsec (center). The AstraLux lucky-imaging approach is then applied, in which a combination of the best 250 images results in a FWHM of 110 milliarcsec (bottom; the high-resolution results for some of the brighter stars are masked by overexposure).
Click here to enlarge image


John Wallace


1. A.S. Fruchter and R.N. Hook, Pub. of the Astronomical Soc. of the Pacific, 144 (February 2002).

More in Optics