Lucky Home Page: Click Here

Figure 2: The High-Resolution LuckyCam on the NTT 3.6m Telescope at La Silla, Chile in June 2006.  Results from this run are described in detail Click Here.

Figure 3and 4: The LuckyCam mounted attached to the Palomar Adaptive Optics unit on the 200 inch telescope in July, 2007.  Preliminary results from this run are described in a series of Press Releases Click Here.

The CCD camera is effectively used in photon counting mode. The gain that we used was typically x2000 giving a signal-to-noise for each detected photon of approximately 7:1. This meant that at low photon rates we are able to do proper photon counting (using methods described in Basden, A. G, Mackay, C.D., et al., (2002), “Low Read-Out Noise CCDs in Optical Interferometry”, SPIE 4838, Hawaii, August 2002). We used a number of different formats to try and explore the capabilities of the technique. Data were taken at frame rates from 100Hz (a tiny field of 100x100 pixels) to 30Hz (512x176 pixels) and the very slow rate of 12 Hz (512x512 pixels giving a field of 20x20 arcsec).

One of the first things we wanted to look at was the size of the isoplanatic patch. We knew that we would measure a smaller isoplanatic patch than the technique will actually allow because the NOT mirror is not diffraction limited, in common with virtually all ground-based telescopes. When using Lucky imaging with these mirrors we are actually waiting for the atmosphere to be as bad as the mirror but with opposite sign. The probability of this happening is not very different from the probability of the phase coming into the telescope being virtually flat. However we would expect to do better with a better mirror.

The movie that can be seen by clicking here is of 100Her, a double star with 14 arc sec separation. 

To see the movie (1.17 MBytes, animated GIF format) click on the above image.

In order to run at a high-speed we electronically  folded the image so that the two components were imaged onto the detector side by side. The scale is about 4 arc sec vertically. The images were taken with 10 msec frame time, and the stars that make up 100Her are each 6.0 magnitude. This sequence makes it clear that even with the separation of 14 arc seconds there is a very high degree of correlation in many frames, particularly when the star images are compact.

The second movie that can be seen by clicking here is of part of the core of the globular cluster M15.

 
To see the movie (12.2 MBytes, animated GIF format) click on the above image.

The brightest stars in this field are about 13.4 Mag, and the faintest visible in each frame are about 15.1 mag. The frame time is 30 msec, and the frame size is 20 x 6.5 arc sec. Again even at this slower frame rate it is clear that when the images are sharp in one part of the frame they are also sharp over much of the frame.

The third movie shown here is also of the core of the globular cluster M15.

To see the movie (15.5 MBytes animated GIF format) click on the above image.

The brightest stars are about  I=13.5 mag, and the faintest visible in each frame are about  I=15.8 mag. The frame time is 80 msec, and the frame size is 20 x 20  arc sec. What is remarkable in this sequence of images is that even at this frame rate which would normally be considered far too slow for any real attempt to freeze atmospheric seeing effects, we still see a remarkable correlation in the good frames where images are most compact over much of the frame.

Using the Lucky imaging methods on a run of 1000 images and selecting 10% of the total we have the image shown here.

There are about 200 stars in this field, and the faintest star easily visible in the field is about  I=17.7 mag (and this limit is set by crowding: in other less crowded parts of the globular cluster,  the limit is I = 18.3 mag.). This image is therefore obtained with 10% of an 80 second exposure or a total exposure time of only eight seconds. The images in this field have a full width at half maximum of approximately 0.13 arc seconds.

In order to allow us to assess image quality properly we need to have an adequate signal-to-noise on the reference object in every selected frame . We can use the data from the globular cluster fields to let us study how faint a reference object we can use reliably.

From these results we find that we can work to ~16.0 Mag (I) with non-thinned CCDs, and not in photon-counting mode. We expect to be able to work with significantly fainter reference stars with thinned  photon-counting CCD. In addition when we are able to use the slowest frame rates such as is shown above in the 512 by 512 taken at 12 frames per second we should be able to reach even fainter limits for our reference star.

At these magnitudes, the  probability of finding a star of a particular magnitude within our isoplanatic patch (50 arcsec diameter) is much greater. That our present limits it is 20% already, and we should reach a very much higher percentage with a thinned CCD , particularly at the lower frame rates. It is also worth remembering that there is nothing to stop us using a mean Strehl ratio of several fainter stars

With Lucky Imaging the highest resolution is achieved by selecting the smallest fraction and therefore only using the very best images. However this gives us a very poor efficiency, and relatively poor limiting magnitude. In practice, however, we can select a larger fraction of images. This will give us somewhat poorer resolution, but still we obtained dramatically better images than we would get from a conventional long-exposure image. Even simply selecting the best 50% of the images means we have chosen the better half, and discarded the worst half. We are able to do this trade-off after the end of the exposure. This allows us to make images by using an increasing fraction of the exposures and deciding whether the improvement in sensitivity is preferable to the slight loss in resolution. The decision about which fraction to use may be made after the exposure, and may be different for different observing programmes, even on the same field.

To give some idea of how this actually works we show here a set of four images each corresponding to a different fraction selected from the same observing run and beside them them a one-dimensional cut through a typical star image.

Plots are for star 10 arcsec from reference star, 12 Hz frame rate and:

           1% selection, 0.13 arcsec FWHM

           3% selection, 0.14 arcsec FWHM

          10% selection, 0.16 arcsec FWHM

          30% selection, 0.18 arcsec FWHM

We find from our results that we already achieve 18.7 mag (I-band) in 80 seconds and 10% selection (i.e. we use 8 secs of recorded data) with an unthinned (front illuminated) L3CCD. The limiting magnitude is potentially very deep.   With 1 hour total exposure and 3% selection, the limit with a thinned CCD would be I~22.3 mag. and would achieve a  resolution of typically 140 marcsec, assuming “seeing” of ~0.5 arcsec. With 10% selection I~23.6 would be reached in about 1 hour, though with poorer resolution (~160 marcsec).

We shall here a composite image. An HST wide field/planetary camera image near the core of M13 makes up the background with Lucky Imaging of one area and Adaptive Optics images of another nearby superposed, along with a comparison of the image profiles of the HST and Lucky Imaging data.

GFDLcontent The work on this page is licensed under the GNU Free Documentation License (http://en.wikipedia.org/wiki/GNU_Free_Documentation_License). The author states that the text and images can be used within the restrictions of this license

Institute of Astronomy & Cavendish Laboratory, University of Cambridge, Madingley Road, Cambridge, UK
© Craig Mackay and the Lucky Imaging Team  2005,2006,2007,2008,2009        .   .
Site last updated on 28 January 2009.    Comments & corrections please, to: cdm @ ast.cam.ac.uk