The L3Vision ccd220 with its ocam test camera for ao applications in Europe

Дата канвертавання24.04.2016
Памер49.15 Kb.

The L3Vision CCD220 with its OCam test camera for AO applications in Europe
Philippe Feautriera, Jean-Luc Gachb, Philippe Balardb, Christian Guillaumec, Mark Downingd, Eric Stadlera, Yves Magnarda, Sandy Denneye, Wolfgang Suskee, Paul Jordene, Patrick Wheelere, Michael Skegge, Peter Poole, Ray Belle, David Burte, Javier Reyesd, Manfred Meyerd, Norbert Hubind, Dietrich Baaded, Markus Kasperd, Robin Arsenaultd, Thierry Fuscof and José Javier Diaz Garciag
aLAOG, Domaine Universitaire, 414 rue de la Piscine, BP 53 38041 Grenoble Cedex 9, France; bLAM, Laboratoire d’Astrophysique de Marseille, 2 place Le Verrier, 13248 Marseille, France;

cOHP, Observatoire de Haute Provence, 04870 St.Michel l'Observatoire, France;

dESO, Karl-Schwarzschild-Strasse 2, 85748 Garching bei München, Germany;

ee2v technologies,106 Waterhouse Lane, Chelmsford, Essex, CM1 2QU, England;

fONERA, BP 72, 92322 Chatillon Cedex, France;

gIAC, Instituto de Astrofisica de Canarias, 38200 La Laguna, Islas Canarias, Spain.


ESO and JRA2 OPTICON have jointly funded e2v technologies to develop a custom CCD for Adaptive Optic Wave Front Sensor (AO WFS) applications. The device, called CCD220, is a compact Peltier-cooled 240x240 pixel frame-transfer 8-output back-illuminated sensor. Using the electron-multiplying technology of L3Vision detectors, the device is designed to achieve sub-electron read noise at frame rates from 25 Hz to 1,500 Hz and dark current lower than 0.01 e-/pixel/frame. The development has many unique features. To obtain high frame rates, multiple EMCCD gain registers and metal buttressing of row clock lines are used. The baseline device is built in standard silicon. In addition, two speculative variants have been built; deep depletion silicon devices to improve red response and devices with an electronic shutter to extend use to Rayleigh and Pulsed Laser Guide Star applications. These are all firsts for L3Vision CCDs.

These CCD220 detectors have now been fabricated by e2v technologies. This paper describes the design of the device, technology trade-offs, and progress to date. A Test Camera, called “OCam”, has been specially designed and built for these sensors. Main features of the OCam camera are extensively described in this paper, together with first light images obtained with the CCD220.

    Keywords: Adaptive optics, AO systems, Electron Multiplying CCD, EMCCD, L3Vision CCD, readout noise, wavefront sensor.


The success of the next generation of ESO (European Southern Observatory ) instruments1 (e.g. VLT SPHERE, MUSE, HAWK-I) for 8 to 10-m class telescopes will depend on the ability of Adaptive Optics (AO) systems to provide excellent image quality and stability. This will be achieved by increasing the sampling and correction of the wave front error in both spatial and time domains. For example, future Shack Hartmann systems will require 40x40 sub-apertures at sampling rates of 1-1.5 kHz as opposed to 14x14 sub-apertures at 500 Hz of current AO systems. Detectors of 240x240 pixels will be required to provide the spatial dynamics of 5-6 pixels per sub-aperture. Higher temporal-spatial sampling implies fewer photons per pixel therefore the need for much lower read noise (<<1e-) and negligible dark current (<< 1e-/pixel/frame) to detect and centroid on a small number of photons. Existing detectors do not have this combination of image area size, read out speed, and low noise performance. As the WFS detector has been identified as the critical component of future AO systems there was an urgency to develop better detectors.

Numerous European astronomy institutions have teamed up in the OPTICON network2, and obtained funds in the Joint Research Activity JRA23, “Fast Detectors for Adaptive Optics”, from the European Commission to support the massive R&D effort to develop a new detector. This development was jointly funded by ESO and OPTICON. After extensive market research culminating in a Call For Tender, e2v technologies was chosen to develop a custom-designed detector based on an extension of their L3Vision5 EMCCD technology. Analysis6 showed that the sub-electron read noise of L3Vision CCDs clearly outperformed classical CCDs even though L3Vision devices exhibits the excess noise factor F of 2 typical of EMCCDs. The reason for this conclusion is clearly shown in the results (Fig. 1) of an analysis performed by Thierry Fusco7 for the ESO instrument SPHERE for two different types of guide stars (GS), white-yellow and red, where a much higher Strehl Ratio is achieved for a faint guide star by an EMCCD than a classical CCD even though it was assumed that the classical CCD had a much higher quantum efficiency in the red.

Fig. 1. Results of analysis performed by Thierry Fusco for ESO instrument SPHERE that compares an EMCCD of read out noise (RON) 0 and 1e- to a classical CCD of read noise 2, 3, and 5e- for two different types of guide stars. Left: Plots of Strehl Ratio versus GS magnitude for white-yellow guide star. Right: Plots of Strehl Ratio versus GS magnitude for red guide star.

e2v has now completed the manufacture of a number of CCD220 devices and the testing of these has started using the OCam camera. Delivery of the science grade devices meeting all requirements are expected in the first half of 2009. After the initial period of detector testing, the CCD220 will be offered to the astronomical community by e2v.


The OPTICON JRA2 science working group set the top level requirements after carefully considering the needs of AO systems for future instruments and their science programs. The following detailed requirements were established:

  • big pixels, square 24 µm (goal), to ease the optical system design, but not too large to produce excessive dark current (DC) or Charge Transfer Efficiency (CTE) problems.

  • versatility of a 100% fill factor and 240x240 square grid array of pixels that can be used by any WFS system: SH, curvature, or pyramid, with or without gaps (guard bands) between sub-apertures.

  • format size of 240 pixels being a number that is divisible by the number of output nodes, 8, and binning factors and aperture sizes of 1, 2, 3, 5, 6 and meets the minimum pixel requirement of 40 sub-apertures x 6 pixels/subaperture.

  • low read noise of < 1 e-/pixel and goal of 0.1 e-/pixel.

  • range of operating frame rates from 25 frames/s (fps) for use when photon starved with faint Natural Guide Star (NGS) to highest sampling rate of 1,300 fps for use with bright NGS and Laser Guide Star (LGS).

  • easy to use; eight output nodes each operating at maximum pixel rate of 15 Mpixel/s, that provide a good compromise between the number of connections from the detector to the outside world and operational practicalities such as power dissipation, pixel rates, and clocking rates.

  • low image smearing (<5%) when transferring from the image to store area; an undesirable effect that can be corrected for by software.

  • cosmetically defect-free as every defect will either complicate the centroiding or make it impossible to centroid a sub-aperture.

  • good spatial characteristics, PSF < 0.9 pixel FWHM over 460 to 950 nm, to accurately determine where the photons are detected.

  • very low Dark Current (DC) of < 0.01 e-/pixel/frame at 1,200 fps and < 0.04 e-/pixel/frame at 25 fps to minimize the large errors introduced by the quantum nature of DC, as the electron is the smallest unit. A single electron of DC creates a large error when centroiding on a small number or single photon event. DC includes contributions from the clock induced charge (fixed amount per frame readout), the image area during exposure (proportional to exposure time), the store section and the serial register during readout (proportional to frame read out time).

  • Peltier cooled package for compact size, maintenance free, and minimal support equipment so that the final assembled camera system can fit in the small volumes usually reserved for AO systems without use of liquid nitrogen cryostats.

  • detection signal limit of 5 ke-. In normal operation, the system will be photon starved as there are not too many bright Natural Guide Stars (NGS) and Laser Guide Stars (LGS) will be operated at as low a power as possible. Well depth and output amplifier dynamic range can be traded to improve other parameters such as higher gain of output amplifier and lower clock amplitudes to transfer charge.

  • Non linearity of < 2%. Analysis shows that this level of non-linearity does not introduce any significant errors. Non-linearity can be corrected by a look-up table.


The CCD220 was the name chosen by e2v technologies for this detector. The CCD2208, 9, 10 (schematic in Fig. 2) is a 24 µm square 240x240 pixels split frame transfer back illuminated L3Vision CCD. The image and store area (store is optically shielded) are built with 2-phase metal-buttressed parallel clock structures to enable fast line shifts in excess of 7 Mlines/s for total transfer time from image to store of 18 µs and low smearing of under 2% at 1,200 fps. Eight electron-multiplying L3Vision4, 5 registers operating at greater than 13 Mpixel/sec enable sub electron noise to be achieved at frame rates of 1,300 fps. With an expected output amplifier read noise of 50e- at unity gain, an electron-multiplying gain of 500 will enable an overall effective read noise of under 0.1 e- (50 e- RON/500 of gain register) to be achieved.


The store area is slanted out to make room for the standard serial registers (three phase clocking) to curve around (Fig. 3) and provide space for the output circuitry. Each output has a 520 element 16 µm standard L3Vision gain register whose gain is controlled by the voltage of the multiplication phase. The output amplifier is a 2-stage source follower and of similar design to that employed on recent L3V CCDs (CCD9711). The gain register is optimized for a gain of 500-600, a value typically expected for AO applications. With an expected output amplifier read out noise (RON) of 50e, this will provide an overall effective read noise of under 0.1 e- (50 e- RON/500 of gain register). The serial registers, gain registers, and output amplifiers are designed to operate up to 15 Mpixel/s to achieve a full goal frame rate of over 1,500 fps. Frame rates exceeding 1300 fps have been achieved so far with 13.6 Mpixel/s pixel rates.

Fig. 3. Details of serial and gain register.

The baseline device is built in standard silicon, which is the low risk option. It is expected that devices of this build standard will be delivered which will meet the minimum requirements. This meets the risk profile of both OPTICON JRA2, who must produce a design report to the European Union by end of 2008, and ESO, who require working detectors for their next generation of instruments. A split wafer run will enable two speculative variants to be built. The first is to build devices in 40 µm, deep depletion, high resistivity silicon (1500 which will offer much better red response (Fig. 4). High red response is important for applications that rely on natural GS such as VLT SPHERE. The second is to build devices with an electronic shutter to extend the use of the detector to applications such as Rayleigh and pulsed Laser Guide Stars (LGS) which require very short shutter times of 2-3 µs. These short shutter times are not possible by mechanical means. Pulsed LGS systems offer the advantage to freeze the laser pulse to a small spot in the sodium layer and thus overcome the problem of spot elongation.

Fig. 4. Comparison of QE for a standard silicon device (basic-90/basic-mid) to that of a deep depletion device (dd basic-90). These curves are only typical and the curves of the final devices may differ.


The CCD220 is encapsulated in a 64 pin package (see Fig. 5) with a custom-designed integral thermo-electric cooler that has been verified11 to cool the CCD below -45°C to achieve the required < 0.01 e-/pix/frame total dark current. The package is sealed and back-filled with 0.9 bar of Krypton gas to minimize heat transfer to the outside.

The CCD220 die has four fiducial crosses, two on either side of the device image area, in the region of the bond pads. These are clearly visible through the package window for alignment purposes such as when attaching lenslet arrays. An AD590 temperature sensor is glued to the ceramic chip carrier to provide a sensor for temperature regulation. The sapphire entrance window is of a high optical quality (double path wavefront error of < 50 nm rms), good surface quality (defects meet 5/2x0,05 DIN3140), and AR coated with transmission > 98% over range 400-950nm.

We performed extensive thermal modelling of the CCD, Peltier cooler, package, proposed clamping arrangement, and water-cooled heat exchanger. The results show that for 10 °C water temperature in the heat exchanger, the Peltier can cool the CCD to below -45 °C. The result of thermal simulation of CCD temperature for different Peltier currents can be found in the right plot of Fig. 6 and is compared to real measurements. This will enable the dark current specification (0.01 e/pix/frame) of the standard silicon device to be easily achieved when the image area is operated in the fully inverted mode and comes very close to meeting the goal requirement (0.001 e/pix/frame). Refer to left plot in Fig. 6 for modelling results provided by e2v technologies of dark current versus temperature.

The high speed operation of the package was optimised by reducing parasitic inductance of the connections between the CCD and the pins of the package by increasing track widths as much as practical and providing ground planes above and below tracks on the ceramic insert (to which the CCD is glued) and package ceramic feed-through. In addition, the camera electronics uses a special pressed contact arrangement to make electrical contact at the ceramic feed-through rather than at the end of the package pin if a traditional ZIF socket was used and to have extremely short connecting path between the amplifier and the CCD pins. Simulations showed that any parasitic inductance forms an LC circuit with the clock phase capacitances (especially the high capacitances of the image and store clocks) and if care is not exercised could generate destructive ringing voltages or when damping serial resistances are inserted reduce the maximum attainable clocking speed.

Fig. 5: Photograph of CCD220 package. The package contains an integral Peltier that has been verified to cool the CCD below -45°C to achieve < 0.01 e-/pix/frame total dark current.

Fig. 6. Results of thermal and dark current modelling. Right: Predicted temperature of the CCD versus Peltier current. Left: Dark current versus temperature modelled by e2v for the baseline standard silicon CCD220 when the image area is operated fully inverted.

5.The OCAM test camera


As part of the OPTICON JRA2 activity, a consortium of French institutions LAOG-LAM-OHP1, has developed a low noise state of the art detector test camera, called OCam13 (photo left Fig. 7) that has been loaned to e2v for testing the CCD220. The image on the right of Fig. 7 was taken by the OCam camera with an engineering CCD220 warm at unity multiplication gain. With a normal camera at same exposure time, i.e. 700 µs, this image was completely black. A copy of the Test Camera will be used by the IAC2 to do acceptance test of the CCDD220. For the scheduled deployment of the CCD220 on VLT instruments, ESO are developing the New General detector Controller14, 15, 16 (NGC). NGC will re-use much of the front-end analogue design developed by the Ocam team (Philippe Balard, Philippe Feautrier, Jean-Luc GACH, Christian Guillaume, and Eric Stadler).


This controller development had a major issue due to the use frequency signals in a range that we are not familiar with (more than 13MHz for most of them). This was a major concern since drive currents are high at the same time due to the fact that CCD behaves electrically like a capacitor. A particular care has been taken to identify the most critical signals necessary to the controller. A good estimate that can be used to identify them is the complex impedance of the load given by :


where F is the equivalent operating frequency in MHz and C the load capacitance. In the case of square wave drives, the equivalent frequency is dependant only on the rise/fall times needed (or the shortest of the two when they are not equal). This equivalent frequency F is given by :


where  is the rise or fall time in seconds.

Another way to find the critical signals is to estimate the instantaneous power dissipation in the load (which is also connected to the impedance computed previously) but also takes into account the voltage swing V across the load:


Where P is the dissipated power, V is the voltage swing and F the frequency.

The Fig. 8 shows the power dissipation in the load vs. frequency. This gives a good estimate of the difficulty of the design which is somewhat proportional to this dissipated power. One can see then that it’s more than ten times harder to generate a parallel phase signal (30MHz equivalent frequency, 1nF load) than an ultra high speed digital design (1GHz, 5pF load). This is due to the fact that a heavy load (low impedance) needs a huge current to be driven at a given frequency. This current is also dependant on the slope needed. When driving these currents at those frequencies, the parasitic inductance is an important parameter since it’s an additional resistance to the current flow and it’s also a resonator in a quasi perfect LC system formed by the inductor and the CCD clocks capacitance.

Fig. 8: Instantaneous dissipated power in the load, gives an estimate of the difficulty to build a suitable driver circuit. Large line is for the parallel phases (1nF load, 8V swing) thin line is for the serial phases (100pF load, 12V swing) and for comparison the dashed line is a classical digital design (5pF load, 3.3v swing). The hatched region is the domain used by the OCAM drivers.

There are several ways to quantify the parasitic inductance introduced by a conductive element. This might vary when the return current path is close to the outgoing current. In fact, it is dependant on the surface of the current loop: the smaller it is, the smaller is the parasitic inductance. For an isolated wire, a good rough estimate is 10nH/cm. In reality it is dependant on the geometry of the wire or track, but the variations are quite small (7 to 12nH/cm in practice). For more complex systems like PCBs, a lot of formulas can be found in the literature, but it is best to use a field solver like hyperlynx to have the exact value.

The transposition of classical design techniques is in this case quite inapplicable since the used frequencies used and loads are rather different than what is usually found with other CCD designs. In our case, the mechanical design has been done around the critical parts in order to avoid adding another level of constraints to the electronics design.

5.2Global architecture of OCAM controller

The controller is divided in 4 parts, see Fig. 9: an acquisition system, an interface, internal microcontroller that manages the drive electronics and the link with the acquisition system.

Fig. 9: OCam controller structure. Left: OCam controller global architecture; (right) OCam controller design with a more detailed breakdown structure of the electronics.
This design is also a low noise design. Therefore a particular care has been taken to minimize RF perturbations. The drive electronics are as close as possible to the CCD in order to minimize the parasitic inductance and allow the use of a higher parallel clocking frequency (see Fig. 10). Only a few centimetres separates the CCD die from the video preamplifiers with the OCam design.

Fig. 10: OCam controller views showing the front end electronics as close as possible to the CCD220.

5.3Main OCAM characteristics

The Ocam system is capable of driving all the CCD220 family CCDs at their nominal speed (1.5kframes/s) and sending all data through a cameralink full interface. The controller might drive deep depleted variants with multilevel clocking at levels up to 24V with a speed of 10Mlines/s (at a nominal phase load of 1nF). It handles the 8 L3vision outputs with high voltage clocking up to 50v. A big effort has been made to have a good high voltage stability (less that 1mv/hour of drift) to ensure a constant gain over a long period. The system digitizes the CCD signal with correlated double sampling with 14 bits dynamics. Interfacing with the camera is quite simple and the actual acquisition system is a PC computer running windows XP fitted with a cameralink full grabber and proprietary software capable of gathering in real time the astonishing 220Mbytes/s produced by the camera.

Also the team has developed a user friendly timer file editor to manage the sequencer of OCAM. The sequencer itself is the heart of the system and has a nominal resolution of 1.5ns and is capable of generating clocks at a frequency of 327MHz. The actual phase jitter was measured at a level of 60ps RMS.

5.4Acquisition software

  • Firmware

In order to fully assess the capabilities of OCAM and of the CCD220 and its variants, a custom set of software applications was designed.

At the lowest level, OCAM runs its firmware on an AVR 8-bit RISC microcontroller, see Fig. 11. The microcontroller is responsible for managing all non-image data input/output of OCAM through the serial lines embedded in the CameraLink standard. The firmware works entirely with text commands so that OCAM can be operated through any system able to talk through RS-232 serial line protocol.

Fig. 11: View of the micro-controller board developed for OCam

  • Timer File Editor

OCAM’s sequencer is fully logic-driven and allows for practically arbitrary clocking of all the phases of the CCD220. In order to ease its operation, a custom timer file editor has been developed to provide a graphical interface over the different clocks used by the detector and to give a powerful tool for the detector testing.

Fig. 12: The OCAM Editor user interface developed for OCam in order to build the clocks necessary to operate the CCD220.

There are 16 General Purpose phases with a time resolution of 9.16ns, 8 High Resolution phases accurate to 1.5ns and 4 High Definition phases that provide up to 16 different levels (with 12-bits precision over the value itself). The Editor is organized hierarchically in Letters, Words then Phrases, each individual item having its own repetition counter. For instance, the exposure time can be set anywhere between 73ns and 237 days (in steps of 73ns). The Editor saves its data in text format, so it can be reviewed and or modified by other applications or directly through any text editor.

OCAM holds 8 different modes (sequencer files) in a non-volatile memory, 5 “standard” modes and 3 user modes.

  • Acquisition Software

The CCD220 and its variants required special acquisition software to cope with its outstanding capabilities, so one was written from the ground up using MS-Visual C++ and Windows XP operating system.

Starting with a standard CameraLink Full Framegrabber (Dalsa-Coreco X64CL-Full), the acquisition software is responsible for grabbing OCAM images at full 1.5KHz speed, providing real-time image analysis tools, saving image data in standard formats (namely FITS and Comma-separated-Values) and of course giving access to all of OCAM commands and values. The acquisition software interface is shown in Fig. 13:

Fig. 13: OCAM Acquisition software interface. Left, top to bottom: Image histogram display, Image slice display. Right, main window of the control panel with Image Area showing the combined descrambled 8 outputs (overscan masked) with data saving options and LUT choices.

All voltages can be set using convenient sliders or numerical edit boxes. Precision over the board is below 0.01V. The “IRD ammeter” shows the Reset Drain Current of the whole image with a precision of 10 pA. OCAM image grabbing is always done in real time as well as image analysis. Image display is done at user-selected frequency anywhere between freeze-frame and 60Hz (screen limit). When examination of successive images is required, a memory buffer can store OCAM video outputs and play back at human speed as well as sum and/or average the images grabbed. Note that this is rather memory hungry as 5 seconds of capture takes a Gigabyte of memory. An option is available to save/load the raw data on disk as well as by using FITS or .csv formats.


As shown in Fig. 14, the CCD 220 first tests are currently starting in the e2v clean room using the OCam test camera. The major performances of the CCD220 with OCam should be known by the end of summer 2008.

Fig. 14: the CCD220 and OCam under test at e2v's clean room (May 2008).


Several European institutions under the umbrella of OPTICON have formed a very good working relationship with e2v to develop a new 240x240 pixel wavefront sensor detector, CCD220, with subelectron read noise at 1,300 fps that will enable future AO systems to provide the image quality and stability to guarantee the success of the next generation of instruments on 8 to 10-m class telescope. Baseline development is low risk by extending existing e2v L3Vision technology to multiple outputs and metal buttressing parallel clock structures. In addition higher risk (more speculative) but higher performance devices in deep depletion silicon and with an electronic shutter will be developed in parallel.

The challenge to obtain good PSF, low dark current and CICs at the same time for the deep depletion variant was described together with operating and clocking techniques that should allow all specifications to be simultaneously met.

For compactness and low maintenance the CCD is mounted in an optimised Peltier cooled package. Modelling and measurements has shown that by using an optimised custom Peltier, an operating temperature lower than -45 °C can be obtained and this will enable dark current specifications to be surpassed.

The test camera, called OCam, is an extremely challenging camera development for testing the CCD220 performances. These tests are currently starting in e2v’s clean room.


  1. ESO telescope and instrumentation website:

  2. G. Gillmore, OPTICON FP6 contract number is RII3-CT-2004-001566.

  3. P. Feautrier,, P. Feautrier et al, “Zero noise wavefront sensor development within the Opticon European network”, Scientific Detectors for Astronomy 2005, pp. 315-320, June 2005.

  4. P. Jorden,

  5. P. Jerram, P. J. Pool, R. Bell, D. J. Burt, S. Bowring, S. Spencer, M. Hazelwood, I. Moody, N. Catlett, and P. S. Heyes, “The LLCCD: low-light imaging without the need for an intensifier”, in Proc. SPIE, 4306, pp. 178-186, May 2001.

  6. T. Fusco et al, “Optimisation of Shack-Hartman based wavefront sensor for XAO systems”, Proc. SPIE Int. Soc. Opt. Eng. 5490, 1155 (2004).

  7. T. Fusco et al, “Shack-Hartmann wavefront sensing with a low-light level CCD: from theory to practice”, proc. SPIE 7015-182, Marseille (2008).

  8. M. Downing, et. al., “Detectors for AO wavefront sensing”, Proc. SPIE 7015-62, Marseille (2008).

  9. M. Downing, et. al., “A Dedicated L3Vision CCD for Adaptive Optics Applications”, Scientific Detectors for Astronomy, 321-328 (2005).

  10. M. Downing, et. al., “Custom CCD for adaptive optics applications”, Proc. SPIE 6276-19, Orlando (2006).

  11. e2v technologies, “CCD97-00 Back Illuminated 2-Phase IMO Series Electron Multiplying CCD Sensor”, Data sheet, A1A-CCD97BI_2P_IMO, Issue 3, May 2004.

  12. P. Feautrier, et. al., “Thermal modeling of cooled instrument: from the WIRCam IR camera to CCD Peltier packages”, Proc. SPIE 6271-41, Orlando (2006).

  13. J-L. Gach et al, “A dedicated controller for Adaptive Optics L3CCD developments”, Scientific Detectors for Astronomy 2005, pp. 639-644, June 2005.

  14. J. Reyes et al, “NGC Front-end for CCDs and AO application”, Scientific Detectors for Astronomy 2005, pp. 579-584, June 2005.

  15. C. Cumani et al, “Software for the New General detector Controller”, Scientific Detectors for Astronomy 2005, pp. 585-588, June 2005.

  16. M. Meyer et al, “NGC Detector Array Controller Based on High Speed Serial”, Scientific Detectors for Astronomy 2005, pp. 571-578, June 2005.

1 LAOG-LAM-OHP, Laboratoire d'Astrophysique de l'Observatoire de Grenoble, Laboratoire d'Astrophysique de Marseille, and Observatoire de Haute Provence.

2 IAC, Instituto de Astrofisica de Canarias, 38200 La Laguna, Islas Canarias, Spain.

База данных защищена авторским правом © 2016
звярнуцца да адміністрацыі

    Галоўная старонка