Written by P.Amico, T Bohm,
Updated and adapted by C.Cavadore

INTRODUCTION

ESO's plan for optical detectors ref[1] are mainly concentrated on commissioningCCDs systems for all the VLT optical instruments, with a total of about40 scientific systems in the next 5 years. These systems will all be drivenby the new FIERA controller ([2], [3]). In addition, all new detectorsmaintained by ESO in La Silla will also be "embedded" in FIERA systems.
While routine testing is usually done at the telescope site, a firstcomplete characterization of each device is carried out at ESO's headquartersin Garching. For example, all the CCDs operating at the La Silla Observatory(12 mounted at the telescope plus about the same amount available for backups)have been tested with the existing ESO testing facility, equipped to operatewith VME and ACE ([4]) controllers.
The same procedure is foreseen for all the detectors needed by theVLT instruments. The introduction of the new FIERA controller as well asthe growing size of CCDs, which cannot be handled by the "old" testingfacility, have made the construction of a new testbench essential for thecharacterization of the future generations of scientific detectors at ESO.
The new testbench prototype for FIERA controlled CCDs, named the TAMeR(Test And Measuring Rig) is at present being developed at ESO/Garchingand has been available since 1998. The new design provides the possibilityto test big detector sizes (8kx8k pixels and bigger) in a one-step process.Fundamental CCD parameters are tested in fully automated modes, while specificadditional tests can be performed interactively. The testbench softwareis integrated with the VLT Standard Central Control Software (CCS, [5])and some of its modules, which will be also available at the telescope,fully comply with VLT standards.
The project is carried out by the ESO Optical Detector Team.

2. TESTBENCH HARDWARE AND DESIGN

The selection of the hardware components has been a lengthy but fruitful process, whose starting point was not only the work done at ESO with theactual testbench facility but also by other laboratories. A useful sourceof information for first step design has been provided by Lesser and McCarthy�swork for the Steward Observatory [6]. A subsequent step has been to createa testbench simulator (by means of Microsoft Excel) to check the actualoutput of the designed system. One of the questions we had to solve was,for instance, whether the employment of a double monochromator, can provideenough flux in the UV range at the detector's position. Although it providesmajor improvement in signal/stray-light ratio with respect to a singlemonochromator, this instrument introduces twice as many reflections, withconsequent loss of signal. The entire system's efficiency has its minimumat the blue end of the spectral range of interest (320 � 1100 nm). Therefore,the selection of many crucial hardware components (precisely the monochromatorand its parts), as well as the verification of small details, such as choiceof the lamp type or the selection of the coating for the integrating sphere,have been taken into consideration and simulated.
A schematic of the final testbench design is given in Fig. 1.

Figure.1: Schematic of the testbench design The elements shown on thetable top (300 X 90 cm) are in scale (except for the dewar size). Fromright to left: lamp and light intensity controller looped via the powersupply, the double monochromator system with two output ports. The secondaryport will input light into the optic fiber, used by PSF experiment; theprimary port feeds the integrating sphere. The flanges are screwed ontothe table top and the space in between light tightened. The CCD head ismounted on a metallic plate fixed to the left flange and the dewar hangsoutside the box. All GPIB driven instruments are connected to the GPIBcontroller, which in turn interfaces via SCSI to the FIERA
 

The system is mounted on a Newport 300 cm x 90 cm optical table top,which provide static rigidity and flatness, in addition to the standardconfiguration of sealed mounting holes. All hardware components are GPIBdriven and controlled through a GPIB controller, by National Instruments,which holds up to 14 GPIB devices and is attached to a SPARC board withembedded FIERA controller through SCSI connection. A PULPO monitoring unit[7] for environmental variable control (temperature, humidity, etc.) willalso be part of the system.
The major system components are:

•  Lamp housing and lamp: a QTH 100 W lamp with horizontally elongatedfilaments is hold in a standard convention cooled housing, equipped witha F/1 condenser, which produce a ~3 cm diameter collimated beam. This beamis then focused on the monochromator input slit by means of a secondaryf/4 plano-convex lens, which matches the acceptance pyramid of the monochromator.The light system maximizes the total power into the monochromator and providesa smooth continuum within the desired wavelength range. The wattage hasbeen chosen in order to have enough flux in the UV part of the spectrum.This condition is especially critical to achieve a good S/N for the absolutecalibrated diode used for QE measurements.
•  Power Supply and Light Intensity Controller: both produced by Oriel.The light intensity controller is directly connected to the lamp housingthrough a light sensing head, which monitors light variations, and interfacedto the power supply. It allows maintenance of constant light levels, forthe duration of an exposure (exposure lengths vary from few seconds toabout ten minutes) regardless of lamp aging, line voltage variations orchanges in the ambient temperature.
•  Monochromator: an Oriel Multispec 257 Double Monochromator in subtractivedispersion configuration. In the current setup, the output from the firstunit is dispersed in the reverse direction by the second unit, thus homogenizingthe light across the output slit. The net dispersion remains as that producedby the first monochromator, but the amount of stray light is greatly reduced,quoted by Oriel to be of the order of 10-7 of the unblocked signal. Thatis, almost three orders of magnitude smaller that the measured stray lightfor a single monochromator of the same kind. The two devices can be controlledvia GPIB in either an independent way or together (using the first oneas master). Both are equipped with microstepping motor driven slits and600 l/mm ruled gratings, whose peak efficiency is at 400 nm. The usablewavelength region (that is, where the grating efficiency is more than 20%),goes from 250 to 1300nm (well beyond our requirements). With this configuration,a minimum bandpass of ~0.1 nm can be reached. Two motorized filter wheels,which hold up to five filters each, are attached at the input of the firstmonochromator. They control respectively the order sorting filters (2 filters,with cut-off wavelength respectively at 450 and 665 nm) and neutral densityfilters (4 filters with optical densities in the range 1-4). The firstunit is also equipped with an integrated shutter, which can be controlledboth via external TTL signals and through GPIB commands. The minimum exposuretime setting is 20msec, the transition time ~2ms. Positioning the shutterbefore the light is inputted into the integrating sphere, instead thatputting it at the exit port, has the advantage of eliminating the shutterpattern problem.
•  Integrating Sphere: a 50 cm diameter "custom made" Labsphere. Its8 inches exit port provides a uniform illumination, over an area biggerthan the size of a 8k × 8k CCD or Mosaic (a typical 8kx 8k with 15um pixel has a diagonal of about 17 cm; we will refer to this example throughoutthe paper). The primary output port is at 180 degrees with respect to theinput port. A secondary output port (about 1.3cm), which hosts a photodiode,is drilled close to the primary output port. A baffle situated inside thesphere prevents that the output port �sees� directly the light source.The internal coating of the sphere is made in Spectraflect, a materialthat ensures a reflectance better than 98% in the range 400 � 1100 nm andbetter than 96% in the UV range. (320-400 nm). The best degree of uniformityacross the illuminated field is achieved when mounting the CCD in closecontact with the exit port. Otherwise, the degree of uniformity, definedas the ratio of the illuminance at the edge of the field to the illuminanceon the axis through the center, is a function of the distance of the targetform the source [8]. The second option has been chosen in order to haveenough space between the sphere exit port and the detector to perform experiments(for instance, to put a lens and a target image to be projected onto theCCDs). The detector will be put at a distance of 50 to 75 cm from the sphereoutput port, so that, for a 8 inches light beam and a 8k X 8k 15 um pixelCCD, the degree of uniformity of illumination is always in the range 95%- 98%. A better than 1% uniformity is of course obtained for smaller detectors.
•  Picommeters and diodes: the testbench will be equipped with two photodiodes,one permanently mounted at the secondary output port of the integratingsphere and the other, needed for absolute flux calibration of the system,put at the detector's position. A permanent solution, with the latter diodefixed as close as possible to the detector and sharing with it the samefocal plane, is also planned for the future. Separate ammeters are attachedto diodes through low-noise triax cables. A Keithley 486 is connected tothe sphere's diode: a 5½-digit autoranging picoammeter designedfor low current applications where fast-reading rates must be performed.The measurement range is between 2nA and 2mA, with a resolution of 10fA(@2nA range). The diode is a Hamamatsu 1cm2 Silicon Photodiode for precisionphotometry (NEP ~10-15) with good UV QE. The second diode is also a 1 cm2silicon Hamamatsu photodiode, which has been absolute calibrated by referenceto NPL (National Physical Laboratory, England) and to PTB (Physikalisch-TechnischeBundesantalt) standards. At present is interfaced to a Keithley 619 Electrometer/Multimeter,with the same measuring range and resolution as the 486 model. Both ammetersare controlled via GPIB by means of the GPIB controller.
•  Flanges system and light tight zone: the integrating sphere is attachedto a flange, fixed onto the table top, through a flexible light shield,which allows a length span of ~25cm. A second flange, which will hold acustom made plate for each detector head (at least three different systemsare foreseen for the VLT detectors systems), is positioned at a distanceof 50 cm from the first flange. The dewar itself will be hanging from theouter wall of the flange. The space in between the two flanges will beclosed by a wooden light-tight box, with lateral access door. The flanges,the box, and some other minor elements are being designed by ESO�s mechanicaldesign office (see figure 1a).

Specifications for the hardware components provided by external manufacturersare listed in the table below:
 

3. TESTBENCH SOFTWARE

>

The requirement to test new generation CCDs driven by the FIERA controllerhas constrained the choice for the software development tools and for thedevelopment platform. The latter must be a SPARC Board with the built-inFIERA system under the Sun Solaris Operating System. The development languageis standard C to allow full compatibility with the FIERA and, more generally,with the VLT software. LabWindows/CVI for Sun Solaris by National Instrumentsis the software development environment of choice, since it offers a fullANSI C and GNU C compatibility. Moreover, applications written under LabWindows/CVIare independent from separate commercial run-time engines (which have tobe licensed). This requirement is essential because the data analysis moduleswill be made available for testing at the telescopes.
In addition, LabWindows/CVI provides very attractive features for buildinginstrumentation software applications: run-time libraries, simplified GUIprogramming, handling of GPIB instrument I/O, use of TCP and a set of instrumentdrivers.
Fig.2 shows a schematic layout for the testbench software. It consistsessentially of a main interface window (GUI), which allows the selectionof the desired testing procedure, either a complete automated sequenceof all main tests or a single user-defined procedure. In all cases, scriptsprovide interactions with three major and independent modules:

1. 1. The hardware control software, which controls all the hardware peripheralsdriven via GPIB. It consists of graphical interfaces plus drivers to allowaccess and manipulation of all the instruments features. Among these, selectionof the wavelength range, of the light intensity, of the bandpass throughmonochromators' control and collection of the data sent from the photodiodesto the ammeters. Flux readings from the ammeters (named control data onFig. 2, because we plan also to use the readings from the sphere's photodiodeto double-check exposure times), setup data from the monochromators andenvironmental data from PULPO are stored for later analysis.

1. 2. The interface module to the FIERA system, which is in charge of sendingto the controller instructions (exposure start, exposure time, I/O operationwith data files, CCD initialization, voltages settings, etc.) for takingsets of exposures. Fits images are stored for later analysis.

1. 3. The Data Analysis module, which takes the data stored by the other modules,process it and extract test results.

Figure 2

Figure 2: the testbench software flow. Different backgrounds distinguishseparate software modules: left) hardware control, which interacts withall GPIB driven hardware; middle) FIERA control, which interact with FIERAsoftware, i.e. with the CCD, and sets exposures sequences and times; right)data analysis, which collects data and process them into a final test report.Also shown is PULPO, used here for environmental control. All modules arecontrolled by a startup GUI, which allows test sequences, parameters definitionand access to the calibration routines

The modules are independent in the sense they exchange data but in generalthey can run in standalone mode. This requirement will allow the use ofthe analysis software at the telescope site, thus providing a standardtool for quick testing of the CCD performances when connected to the instrument.
The main interface accesses also a calibration routines module, usedperiodically to check the system's hardware responses and perform wavelength/fluxcalibrations.
 

figure 3, main testbench's software panels, left side to set andstart test sequences, right to set up (manually) the CCD illumination.

4. TEST PROCEDURES

Once the data have been acquired, the FITS files are moved to a powerfullPC (July-99) Dual PII 450MHZ processor, 512Mb RAM and processed by the  Prism   software. This softwareincludes all the usual routines to get the CCD characterization parameters.
 
 

A standard report is made accordingly to the measurments, and the CCDgrade is rated according this measurmentrs for its future purpose..
This set up was alsoused to generate this EEV 44 report.

5. CONCLUSIONS

Design and hardware implementation phases have been completed and thetestbench has been ran since August 1998. This set up has been used to testall the CCD for the various instrument (FORS1/2,UVES, WFI,...). All the parameters are measured with the accuracyexcepted and, this testbench is extremely useful to characterize CCDs,monitor thier behavior according to time.
At that time, we are improving software tools to acquire the Data,so as to end up with a homogenious software architecture. Also we planto make tests with differents voltages so as to get the best performancepossible, in an automated way.

Acknowledgments: all members of the Optical detector Team, both in Garchingand Chile, have contributed to design and critique the development stagesof the project. We thank M. Lesser for sharing with us his experience inbuilding a test facility. We are also grateful to Fernando Pedichini, whogreatly contributed to the project during his visits to ESO and who isstill actively collaborating with us.

6. REFERENCES

[1] J.W Beletic, these Proceedings
[2] J.W. Beletic, R. Gerdes and R C. DuVarney, these Proceedings
[3] C. Cumani & R. Donaldson, these Proceedings
[4] Reiss R., "ACE, ESO's next generation of CCD Controllers for theVLT", in "Instrumentation in Astronomy VIII, 13-14 March 1994, Kona, Hawaii",SPIE Proceedings, vol. 2198
[5] Central Control Software (CCS0 User Manual, VLT-MAN-ESO-17210-0619
[6] M.P. Lesser and B. L. McCarthy, 1996, Proc. SPIE 2654B, �QE Characterizationof Scientific CCDs�
[7] N. Haddad, P. Sinclaire, J. Anguita, A. R., these Proceedings
[8] R. Kingslake, Illumination in Optical Images, Applied Optics andOptical Engineering, Ed. R. Kingslake, Vol. II, Ch. 5, Academic Press,1965.

See also :

ESO's new CCD testbench, p95, Astrophysic ans space Library, Optical Detector for astronomy, James Beletic, Paola Amico, KLUWER ACADEMIC PUBLISHERS