AstraLux - the Calar Alto 2.2-m telescope Lucky Imaging camera

a r X i v :0807.0504v 1 [a s t r o -p h ] 3 J u l 2008AstraLux -the Calar Alto 2.2-m telescope Lucky Imaging camera

F.Hormuth,W.Brandner,S.Hippler,Th.Henning Max-Planck-Institute for Astronomy,K¨o nigstuhl 17,69117Heidelberg,Germany E-mail:hormuth@mpia.de Abstract.AstraLux is a Lucky Imaging camera for the Calar Alto 2.2-m telescope,based on an electron-multiplying high speed CCD.By selecting only the best 1-10%of several thousand short exposure frames,AstraLux provides nearly di?raction limited imaging capabilities in the SDSS i’and z’?lters over a ?eld of view of 24×24arcseconds.By choosing commercially available components wherever possible,the instrument could be built in short time and at comparably low cost.We brie?y present the instrument design,the data reduction pipeline,and summarise the performance and characteristics 1.1.Introduction The recovery of the full theoretical resolution of large ground-based optical telescopes in the presence of atmospheric turbulence has been a major goal of technological developments in the ?eld of astronomical instrumentation in the past decades.The de-facto standard nowadays is adaptive optics,the active real-time correction of the incoming distorted wavefronts by means of deformable mirrors.While di?raction limited image quality can be routinely achieved in the near infrared at large telescopes,this is only possible with considerable technical e?ort.Besides speckle techniques in the Fourier domain,a relatively simple approach is the selection of images based on their quality:instead of using all images of a large set of short-time exposures,only those showing little image distortion due to the variable strength of atmospheric turbulence are combined to a high-resolution result.Like in the speckle interferometric approach,this requires only a detector with fast readout capability and moderate computational e?ort.The

application of this technique to fainter astronomical targets,e.g.low-mass double stars,was limited by the readout noise of the available detectors at visible wavelengths until a few years ago.At typical exposure times of few ten milliseconds –necessary to “freeze”the e?ects of atmospheric turbulence –photon noise limited detectors are obligatory to use this so-called “Lucky Imaging”technique on targets fainter than ≈10mag.Since the image quality of each single frame has to be determined,e.g.by measuring the Strehl ratio of a reference star,the readout noise of the detector sets stringent limits on the minimum brightness of this reference.The advent of electron multiplying CCDs (EMCCD)with single-photon detection capabilities led to considerable interest in the Lucky Imaging technique on the part of professional astronomers [1,2,3].First experiments at the Nordic Optical Telescope (NOT)with LuckyCam 1

Based on observations collected at the Centro Astron′o mico Hispano Alem′a n (CAHA)at Calar Alto,operated jointly by the Max-Planck Institut f¨u r Astronomie and the Instituto de Astrof ′?sica de Andaluc ′?a (CSIC).

Figure1.AstraLux at the Calar Alto2.2-m telescope telescope.The yellow box contains the 8-position?lterwheel,the camera is attached at the bottom.The grey rack to the right houses the camera control computer,a keyboard/monitor combination,the?lterwheel electronics,and MicroLux,a GPS timing add-on.

proved that Lucky Imaging is a very promising alternative to adaptive optics,allowing di?raction limited imaging at telescopes in the2–3m class at wavelengths below1μm[4,5,6,7].

These results triggered the development of a similar instrument for the2.2-m telescope at Calar Alto by our group.The Lucky Imaging camera AstraLux was built in less than5months thanks to the availability of most parts as o?-the-shelf equipment.

In the following we describe the instrument design and data processing pipeline and summarise the instrument’s performance and key characteristics.Interested readers looking for more comprehensive information are referred to the diploma thesis,which covers AstraLux and its performance in full detail[8]2.

2.Instrument Design

2.1.Telescope and Site Characteristics

The Calar Alto observatory is located at a height of2150m in the Sierra de los Filabres, approximately50km north of Almeria in Andalusia,southern Spain.The median seeing is 0.′′9,slightly better in summer than in winter[9].

Together with the given telescope diameter of2.2m,the seeing de?nes the optimal wavelength for Lucky Imaging:the probability for the occurence of a good frame(i.e.with a Strehl ratio >37%)in a series of short exposure time images decreases exponentially[10]with the ratio of telescope diameter over the Fried parameter r0,which again depends inversely on the seeing and scales withλ5/6.While observations at short wavelengths result in too low probabilities for usable frames,the gain in resolution by selecting only high-quality frames becomes small at long wavelengths due to the decreasing theoretical resolution of the telescope.For the Calar Alto 2.2-m telescope,the optimal wavelength for e?cient Lucky Imaging lies in the range of800–1000nm,including the Johnson I and the Sloan Digital Sky Survey(SDSS)i′and z′passbands. At typical seeing conditions one can expect that≈0.5–1%of all images will have an acceptable Strehl ratio of few10%while providing a6to10times better resolution.

2Available online at http://www.mpia.de/ASTRALUX/

2.2.Camera&Optics

The AstraLux camera head is an Andor DV887-UVB model from Andor Technologies,Belfast, Northern Ireland.It is an electron-multiplying,thinned,and back-illuminated512×512pixel CCD that comes as complete package with a multi-stage Peltier cooler,mechanical shutter, computer interface,and software.

Starting at room temperature,the typical operating temperature of?75?C is reached within less than10minutes.The camera requires neither re?ll of liquid coolants nor any action to maintain the vacuum inside the CCD head.The quantum e?ciency(QE)of the E2V CCD97 detector peaks at>90%in the R band.At912nm,the central wavelength of the SDSS z′?lter, the QE is still≈40%.For Lucky Imaging we use a readout clock of10MHz,giving a frame rate of≈34Hz at an A/D resolution of14Bit.The readout noise is≈80e?.At an electron gain of 2500,this corresponds to a SNR of>30for single photons.

At the Calar Alto 2.2m telescope,the camera’s physical pixel size of16μm roughly corresponds to almost twice the size of the theoretical PSF at912nm(SDSS z′band).For the?nal design,a pixel scale of47mas/px was adopted as a good compromise between spatial sampling and the resulting size of the?eld of view which is24×24′′.The hardware realisation chosen for the AstraLux instrument is a single negative achromat in Barlow con?guration,i.e. placed in the optical path before the nominal focal plane of the telescope.

The instrument’s?lter wheel with8positions can hold virtually any?lter that is available at the observatory3,allowing observations at a wide range of wavelengths.The standard?lters used for Lucky Imaging observations are SDSS i′and z′band interference?lters,manufactured by Asahi Spectra Co.,Ltd.,Tokyo,Japan.The transmission curves are very close to the original SDSS?lter system de?nition[11]and peak at more than95%.

2.3.Instrument Control&Data Processing

AstraLux can be controlled from virtually any point of the observatory with a100MBit Ethernet network connection.The camera control software o?ers a real time display and allows setting of all crucial parameters like exposure time,electron gain,or number of requested frames.The raw FITS data cubes(typically18MB/s at maximum frame rate)are not stored on the camera computer itself,but on a fast RAID-0array with1TB capacity in a gateway machine.

The AstraLux data reduction software is running on a dedicated pipeline computer,equipped with two dual-core Woodcrest processors and8GB of memory.The pipeline automatically produces quicklook results of the Lucky Imaging observations in approximately the same time that is needed for data acquisition.The basic pipeline algorithms are similar to that adopted by the LuckyCam team[12,13].

After completion of a Lucky Imaging observation,the position of a suitable reference object for quality assessment has to be determined.This is performed on a stacked image of the?rst2 seconds of raw data,and can be done either manually or automatically.The pipeline tracks the reference on all following raw data frames to account for large atmospheric tip/tilt or telescope tracking errors.Subsequently,the quality of each single frame is determined by measuring the Strehl ratio of the reference object after extracting,resampling,and noise-?ltering a small region around the reference.

The pipeline’s image reconstruction module performs data reduction in its literal sense.From typically several GB of input data,just a few MB of pipeline results are produced.The most interesting ones–the Lucky Imaging results–are currently generated with the Drizzle algorithm [14].This linear reconstruction method is?ux preserving and able to at least partially overcome the slight undersampling that is present in the raw data.It is capable of handling sub-pixel translations without the need to perform image shifting in the Fourier domain.

3See http://www.caha.es//CAHA/Instruments/filterlist.html for a list of all available?lters.

0.04 0.06

0.08

0.1

0.12

0.14 0.16 0.18

0.2

0.22

0.24

S t r e h l R a t i o Reference star I-band magnitude Figure 2.Dependency of the ?nal Strehl ratio

on natural seeing and reference star magnitude.

The plot is based on observations with 30ms

single frame exposure time in the SDSS z’?lter

and 1%image selection rate. 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4

5M a g n i t u d e d i f f e r e n c e Distance from primary / arcsec

5σ peak detection SDSS z’5% selection Seeing 0.7"Seeing 0.8"Seeing 1.0"Figure 3.Typical achievable magnitude di?erences for a 5σpeak detection of a fainter companion to the reference star.

1??

Figure 4.AstraLux ?rst light:the binary WDS 14139+2906with an angular separation of 0.′′52and magnitudes V =7.5and V =7.6mag,and the 0.′′23separated binary WDS 15420+0027with V =8.2and V =8.8mag brightnesses.Both images are based on a 2%-selection from 10000single frames in the SDSS z’?lter with 30ms exposure time each.Image scaling is

linear.

The drizzling process oversamples the input data twice,resulting in a pixel scale of ≈23.7mas/px in the ?nal images.Currently the pipeline produces drizzled results of the best 1,2.5,5,and 10%of the input frames.Bias and ?at?eld calibration frames are applied to the input images prior to drizzling.A seeing (and tracking)limited image is generated as well using all frames to allow quick measurements of the seeing conditions,useful at times when the observatory’s seeing monitor is switched o?,or for later assessment of the data quality.

3.System Performance

First light at the Calar Alto 2.2-m telescope was obtained on July 6,2006with photometric sky conditions and V -band seeing values as low as 0.′′6.

Known bright double stars were chosen as ?rst light targets and observed in the SDSS z’band with resulting Strehl ratios of ≈20%(see Figure 4).The point spread function was found to show a remarkable radial symmetry,with a typical FWHM of the PSF core of 110mas in the SDSS z’?lter.

Operating the instrument,and especially acquiring targets,proved to be much easier than anticipated.Though the pointing accuracy of the telescope is in general not better than 10′′,the availability of the camera’s real time display allows short acquisition times of typically 1–2min per target.

AstraLux observations of globular cluster centres enabled the characterisation of the image

Figure https://www.wendangku.net/doc/dc5137629.html,parison between seeing limited imaging and the“Lucky”version:The combination of the best5%of10000single frames provided a Strehl ratio of20%in this observation of the core of the globular cluster M15.Though the conventional result contains20 times more photons,it is clearly inferior in terms of point source detection limits.

quality over the full?eld of view.Choosing di?erent stars with a wide range of magnitudes as the Lucky Imaging reference allowed to estimate brightness limits for the reference selection and to measure the dependency of the Strehl ratio on the reference magnitude.Among the globular clusters M3,M13,and M15,the latter has been observed most extensively with AstraLux. Figure5shows a comparison of a Lucky Imaging result to the corresponding seeing limited image.

3.1.Isoplanatic Angle,Coherence Time&Limiting Magnitudes

The observations of centres of the globular clusters M13and M15allowed measurements of the Strehl ratio for a large number of stars,well distributed over the?eld of view.The values were normalised by the Strehl ratio of the reference star,and the isoplanatic angle was determined by?nding the angular separation from the reference where the Strehl ratio drops to1/e.This procedure was applied to observations with di?erent exposure times ranging from15to60ms, consistently resulting in an estimate of the isoplanatic angle of≈40′′in the SDSS z’band.

To determine the speckle pattern coherence time,a series of10000images of the bright star βAnd was recorded with a time resolution of4.6ms.The measured coherence timeτe=36ms corresponds to the1/e point of a Lorentzian?t to the temporal autocorrelation function of the focal plane intensity at a?xed point.

The globular cluster observations were re-analysed with di?erent choices of the Lucky Imaging reference star to assess the impact of the reference magnitude on the?nal Strehl ratio.Figure2 shows the results for measurements under two di?erent seeing conditions.While reference stars as faint as I=15.5mag still allow a substantial improvement of image quality under a0.′′65seeing, the same performance cannot be reached with stars fainter than13.5mag in0.′′85seeing.

The achievable contrast ratio for close companions to bright host stars was determined on?nal pipeline results of SDSS z′band observations of single stars under di?erent seeing conditions. The deduced maximum magnitude di?erences for a5σpeak detection are based on measurements of the noise in concentric rings around these stars.Simulations with observed PSFs were carried out to check the reliability of the numerical results.Figure3shows typical detection limit plots for three di?erent V-band seeing values.At angular separations larger than2′′,the detection limit is determined by readout noise and the Poisson noise of the sky background.

4.Conclusions

Within less than one year it was possible to design,build,and characterise a Lucky Imaging instrument for the Calar Alto2.2-m telescope.As a common user instrument from2007 on,AstraLux has become the standard tool for di?raction limited imaging at the Calar Alto observatory.It is currently mostly used for binarity surveys among stars and minor planets.

AstraLux is able to reach Strehl ratios as high as25%in the z’band.In general,Lucky Imaging provides an improvement of the Strehl ratio by a factor of10,corresponding to an increase of the signal-to-noise ratio for point sources by a factor of10?20,depending on atmospheric conditions.Thus a selection of only the best5?10%of all images does de?nitely not have a negative e?ect on the detection limit for point sources.

The requirements for the reference star magnitude are similar as for observations with adaptive optics.The performance starts to signi?cantly decrease at I=14mag,but image quality improvements are still possible with stars as faint as15?16mag.The measured isoplanatic angle in the z′-band is with≈40′′as large as typical values in K-band for adaptive optics observations.

The measured close companion detection limit at an angular separation of1′′is on average 6mag,worse than what adaptive optics can provide.But:adaptive optics at8-m class telescopes currently has this capability only in the H and K-band at wavelengths>1.5μm.The achievable contrast ratio in speckle imaging observations is typically two magnitudes less than for Lucky Imaging.

The encouraging results regarding both development speed as well as scienti?c output triggered the start of the project“AstraLux Sur”,an almost identical copy of the Calar Alto version to be used as visitor instrument at ESO’s New Technology Telescope(NTT)at La Silla, Chile.AstraLux Sur is currently on its way to Chile after a development time of less than2 months and will have?rst light in mid-July2008.

References

[1]Jerram P,Pool P J,Bell R,Burt D J,Bowring S,Spencer S,Hazelwood M,Moody I,Catlett N and Heyes

P S2001Proc.SPIE4306178

[2]Hynecek J and Nishiwaki T2002Proc.16th World Multiconference on Systems and Cybernetics

[3]Mackay C D,Tubbs R N,Bell R,Burt D J,Jerram P and Moody I2001Proc.SPIE4306289

[4]Baldwin J E,Tubbs R N,Cox G C,Mackay C D,Wilson R W and Andersen M I2001A&A368L1

[5]Tubbs R N,Baldwin J E,Mackay C D and Cox G C2002A&A387L21

[6]Tubbs R N,Baldwin J E and Mackay C D2003Proc.SPIE48391093

[7]Law N M,Mackay C D and Baldwin J E2006A&A446739

[8]Hormuth F2007AstraLux:High Angular Resolution Astronomy with an Electron Multiplying CCD Diploma

thesis,University of Heidelberg

[9]S′a nchez S F,Aceituno J,Thiele U,P′e rez-Ram′?rez D and Alves J2007PASP1191186

[10]Fried D L1965JOSA551427

[11]Fukugita M,Ichikawa T,Gunn J E,Doi M,Shimasaku K and Schneider D P1996AJ1111748

[12]Tubbs R N2003Lucky Exposures:Di?raction Limited Astronomical Imaging Through the Atmosphere Ph.D.

thesis,Cambridge University

[13]Law N M2006Lucky Imaging:Di?raction-Limited Astronomy from the Ground in the Visible Ph.D.thesis,

Cambridge University

[14]Fruchter A S and Hook R N2002PASP114144