The "Bonn University Simultaneous CAmera" (BUSCA) is a CCD camera system which allows simultaneous direct imaging of the same sky area in four colors. The optics are designed for an f/8 beam and four 4Kx4K CCDs with 15m pixels covering a field of view of 12 arcmin x 12 arcmin at a 2m class telescope. In September 1998 BUSCA has seen "First Light".

The instrument is based on three dichroic beam splitters which separate optical wavelength bands such (at 430nm, 540nm, 730nm) that standard astronomical intermediate-band filter systems can be used. The dichroics are made of plane-parallel glass plates mounted at an angle of 45 degrees. Astigmatism in the transmitted beams (f/8) is completely cancelled by identical plane-parallel glass plates of suitable orientation.

BUSCA offers new perspectives in astronomical multicolor photometry: i) The broadband spectral properties (e.g. color indices) of astronomical objects in the optical can be determined with high reliability even in non-photometric atmospheric conditions. ii) Precious observing time is used very efficiently. iii) With the large field of view, extended objects like globular and open star clusters or galaxies are covered in a single exposure. iv) Each exposure results in a complete data set.

Keywords: Astronomy, CCD, large format imager, simultaneous multicolor imaging
 
 

At astronomical optical telescopes CCD camera systems are used mainly as imaging precision photometers. Until the CCD became the detector of choice in astronomy, imaging (photographic) and precision photometry (photoelectric, non-imaging) had been different tasks. They were almost mutually exclusive. Since then the situation has changed significantly. While intrinsically the CCD detector has its imaging capability, these images can be calibrated with very high precision over a large dynamic range. In a single CCD image, photometry of a large number of objects (pointlike or extended), of reference objects and of the background brightness is performed at once. Consequently CCDs have not only almost completely replaced the photographic plate but to a large extent also photoelectric diaphragm photometers. Exceptions are the photometry of rapidly varying sources and/or simultaneous multicolor photometry.

The main aim of astronomical CCD photometry is to investigate with full spatial resolution spectral properties (e.g. characterized by color indices) and spectral features at low or intermediate spectral resolution. Hence it should be considered as spatially resolved multicolor photometry. It is performed in well defined wavelength bands, known as photometric systems or filter systems. Various standard filter systems are used with bandwidths between 100nm (e.g. the Johnson/Cousins UBV(RI)_C system) and 20nm (e.g. the Strömgren uvby system). For narrow spectral features correspondingly narrow filters are used together with at least one additional filter at a neighboring wavelength for reference.

The standard observing technique for multicolor CCD photometry is to get images through all filters sequentially. This has some obvious drawbacks.

• While observing at a given wavelength band the light from all other wanted bands is lost.
• Sequential photometry needs a highly stable atmospheric transparency (so called photometric conditions). Any unrecognized change in atmospheric transparency would result in wrong color indices. Unrecognized non-photometric conditions are the most important and very often the most frustrating aspect of photometric observations.
• The spectrally resolved variability of objects can only be investigated on long timescales (hours).

When it was decided to build a multicolor CCD camera system for the 2.2m and 3.5m telescopes on Calar Alto observatory the main considerations were:

• Observations would be possible in all color channels simultaneously. The Calar Alto Observatory - like most other sites - suffers from changing atmospheric conditions. Nevertheless photometry as one of the principal information provider of observational astronomy plays an important role. As an example, CADIS^6. (for: Calar Alto Deep Imaging Survey), one of the key projects is a 10-color survey project carried out in the traditional sequential mode.
• Standard astronomical filter systems would be used. Requirements were restricted to intermediate band non-overlapping filter systems. The Strömgren uvby filter system would be the first available.
• Prominent emission lines would be accessible. The investigation of galactic and extragalactic emission line regions (star forming regions, planetary nebulae) are part of the astronomical program.
• A reasonably large field of view would be available. The 2.2m RC-type telescope has a maximum usable field diameter of about 90mm (without corrector lens). During the initial design phase mosaics of 2x2 CCDs with 2Kx2K 15m pixel were foreseen. The images should show no vignetting for beams as fast as f/8 (the aperture ratio of the 2.2m telescope).
• For practical reasons (e.g. telescope fork clearance, instrument weight) no reimaging optics would be used and the number of color channels was restricted to four. Still the full optical range would be accessible.
• As the shutter operates in the convergent beam a high degree of homogeneity across the full shutter aperture would be required. The design goals were remaining inhomogeneities of 1 percent or less at an exposure time of 0.1sec - corresponding to a timing accuracy of 1msec.
• Although an optimized CCD controller would have been an option it was a basic requirement by the observatory to follow their controller concept.

2. INSTRUMENT OVERVIEW

1. The "cube" (Fig. 1) This is the heart of BUSCA. A compact design was achieved with a size along the edges of 350mm to 400mm. It contains all optical and optomechanical components including one filter wheel for each color channel.
2. The CCD units (Fig. 1). The 4 nitrogen bath dewars were designed and built in Bonn. They are mounted by means of a special flange that allows easy handling by a single person.
3. The CCD controller. We have assembled a single controller unit which was designed at the Max Planck Institute for Astronomy (Heidelberg/Germany) (MPIA) and is now able to drive 4 CCDs.
4. The instrument control unit. A set of microcontroller systems connected by a serial field bus performs all necessary functions: Positioning of the four filter wheels, CCD temperature control, control of liquid nitrogen evaporation and communication with the workstation.
5. The data acquisition workstation. It is connected with the CCD controller and the instrument control unit. Data acquisition, communication and the user interface for the observer run on this machine.

Fig. 1 BUSCA 3D design
This drawing shows the central "cube" and the four CCD dewars. The telescope flange is at the top.

The "cube" contains the beam splitter unit, the shutter and in front of each dewar a filter wheel with four positions.

3.1 The beam splitter unit

Astronomical multicolor photometry is based on a number of different more or less well defined filter sets. Among the most widely used are the Johnson/Cousins UBV(RI)_C system with broad (about 100nm) overlapping filter transmission curves and the Strömgren uvby system with intermediate band (20nm – 40nm) filters which do not overlap. Moreover a large number of narrow band line filters are used for the spatial mapping of emission line objects. Although simultaneous multicolor photometry with overlapping filter transmission curves may in principle be possible using broad band beam splitters with the full spectrum available in each channel, such a system would use light with very low efficiency. A combination of dichroic beamsplitters and intermediate and/or narrow band filters appears to be the optimum.

Although good spectral resolution and a large number of spectral samples would be advantageous also for multicolor photometry we have decided to build a system with four color channels for scientific and practical reasons. The primary research topic of BUSCA will be the study of global physical parameters of stars and stellar systems and the investigation of the warm ionized interstellar medium with a reasonable field of view. The four intermediate band filters of the Strömgren photometric system are particularly well adapted to stellar research. The primary target telescope for BUSCA is the 2.2m telescope at the Calar Alto observatory. The restrictions set by the telescope back focal distance on the one hand and the dewar size for large format CCDs on the other hand led to a system which was limited to a maximum of four color channels.

The BUSCA beam splitter consists of 3 dichroic plane parallel plates which all operate at an angle of 45 degrees. The primary plate operates at l @ 540nm splitting the light into a reflected blue and a transmitted red secondary beam. The two secondary plates split the blue beam at l@ 430nm and the transmitted red beam at l@ 730nm respectively. This results in the tertiary 4 wavelength bands or color channels ranging from the UV to the near IR. The dichroic plates are made of BK7 glass (thickness: 10mm) and their second surfaces have got an appropriate anti reflection (AR) coating. Fig. 2a shows the arrangement of the beam splitter plates in a schematic drawing and Fig. 2b the four passbands based on the manufacturers spectral scans of individual plates. For the four BUSCA bands the following designations will be used: uv_B, b_B, r_B, nir_B. Without additional filters the CCD QE curves define the lower and upper wavelength boundaries in the uv_B and nir_B bands respectively. The horizontal bars in Fig. 2b indicate the wavelength ranges covered by filters of common photometric systems (Strömgren, Thuan/Gunn, Johnson/Cousins). The Thuan/Gunn system is mainly used for studies of external galaxies. The band edges were defined as a compromise between simultaneous availability of a maximum number of filters of a system and having bands with comparable widths over the full optical range. The nir_B band was chosen so that the Johnson/Cousins I_C Filter fits into it. The predominance of filter bands on the blue side of the spectrum which is typical for many classical photometric systems is to a large extent due to the fact that photographic emulsions and photomultipliers have no or very low sensitivity in the NIR.
 

Fig. 2 Dichroic beamsplitter characteristics 
a) Schematic drawing of the relative orientation and position of the dichroic plates, filters and CCDs along the light paths. b) Effective transmission of the four wavelength bands. The transmission curves are based on spectral scans of the individual plates by the manufacturer. The two series of 4 horizontal bars designated u, v, b, y and u, v, g, r indicate the Strömgren and Thuan/Gunn filter systems. IC is from the Johnson/Cousins system.

The use of an inclined plate in a convergent beam is optically disadvantageous. It produces strong astigmatism in the transmitted beams. But this solution was dictated by the pure size and weight of massive blocks of glass which would have been the alternative (not to mention the introduction of huge amounts of spherochromatism). Light in the uv_B channel experiences only reflections (two) and is therefore unaffected. Light in the b_B and r_B bands is transmitted once and in the nir_B band twice. For the latter case a very simple solution was found. Instead of mounting these two plates with parallel or antiparallel orientations as indicated schematically in Fig. 2a, the second plate was rotated by 90 degrees around the optical axis (see Fig. 3). Ray tracing simulations are shown in Fig. 4 for both situations, a single plate and two plates respectively. The astigmatism which is dominating in Fig. 4a is completely cancelled by introducing the second rotated plate. Two residual aberration effects on a much smaller scale appear in Fig. 4b: Coma which is parallel to the plate diagonal and transverse chromatic aberration. But both effects are of the order of a single CCD pixel (15m) which corresponds to 0.17 arcsec at the 2.2m telescope and is in any case small compared to a typical seeing condition of 1 arcsec. In order to also compensate astigmatism in the b_B and r_B bands we have added glass plates of the proper inclination and orientation to both channels (see Fig. 3). These plates have only AR coatings.

Tests of the optical performance of the complete beam splitter unit consisting of five plane parallel plates were made on the optical bench with a star test configuration. A mirror system consisting basically of two Newtonian telescopes facing each other was set up such that an outgoing beam with f/8 and a sufficiently large back focal distance (400mm) was achieved. The target, a 15m pin hole, was enlarged to 20m, still comparable to the size of the CCD pixels of 15m. The CCD detectors were 2Kx2K devices from LORAL. Each channel had the CCD system we have been using later at the telescope during the test runs. The CCD at the uv_B channel was a backside thinned device (by Mike Lesser, Steward Observatory; PPtF technology). The remaining three were thick devices, one of them (b_B channel) with a METACHROMEÒ  II coating (by Photometrics, Tucson). The point spread functions (PSF) measured for each channel are shown in Fig. 5. We find as a main result, that all PSFs are symmetric and the full width at half maximum (FWHM) is between 1.9 and 3.1 pixels, corresponding to 0.33 and 0.54 arcsec at the 2.2m telescope. Measurements for the b_B and r_B channels (2.0 and 1.9 pixels) may even have suffered from undersampling which would make those values upper limits. The FWHMs of the uv_B and nir_B channels are significantly larger (2.7 and 3.1 pixels) than those of the two intermediate channels. Although effects from the beam splitter plates can not be completely excluded we interpret these results as MTF degradation of the CCDs known as diffusion MTF⁷ originating in both cases from the wavelength dependence of the absorption length in silicon. It is well known that backside thinned CCDs show a MTF degradation at shorter wavelength due to electron diffusion as electrons are produced close to the surface in the field free region. On the other hand NIR photons entering a thick CCD from the front side can cross over the depletion region due to their larger absorption length (> 10m) and will produce their electrons in the field free region too. Our results are compatible with calculations of the NIR diffusion MTF⁷ for thick CCDs and measurements of the MTF of LORAL/Lesser CCDs^8,9.

Fig. 3 Light paths inside the “cube”.
This 3D view shows the placement of the three dichroic plates (light gray) and the two additional corrector plates (dark gray) inside the housing, and the four filter wheel units with the CCD dewars connected to them. All plates have an inclination angle of 45 degrees with respect to the optical axis. The individual light paths are indicated. Light comes from the top and is split at the primary 540nm-dichroic. The uv_B light is there reflected backward and sideward to the left at a secondary 430nm-dichroic. The b_B light transmitted at this secondary dichroic needs to pass a corrector plate. An additional corrector plate is also needed for the r_B light which is transmitted at the primary and reflected at the 730nm-secondary plate. The nir_B light is transmitted through two dichroics which have the proper orientation for a correction. 

Fig. 4 Astigmatism correction
a) Intra and extra focal spot diagrams for an astigmatic f/8 beam at 550nm transmitted through a 10mm thick plane parallel plate (BK7) which is inclined by 45 degrees. All units are millimeters if not otherwise indicated. Numbers at the spot diagrams give the distance from the circle of least confusion. The open circle in the center of each spot diagram is the airy disk (approximately 13.5m).
b) Spot diagrams for a compensated f/8 beam transmitted through 2 10mm thick plane parallel plates (BK7) which are inclined by 45 degrees and rotated relative to each other by 90 degrees around the optical axis. Diagrams are shown for three wavelengths across the Strömgren b band, were residual effects are expected to be largest. Residual coma and chromatic transverse aberration can be recognized.

The residual optical aberrations seen in our test measurements, in particular the transverse chromatic aberration, would be strongest in the b_B and nir_B channels: in the b_B channel because the differential dispersion across the Strömgren b band is larger than across the Strömgren y band in the r_B channel; in the nir_B channel because the band was not narrowed by an additional filter. But we should then see PSFs which are elongated along an angle of 45 degrees (see Fig. 4). The PSF intensity maps in Fig. 5 with logarithmically enhanced low intensity levels show that this is not the case. This is as expected from the ray tracing experiments where the amount of transverse aberrations is of the order of a single pixel.
 
 

Fig. 5 Point spread functions (PSF) of individual color channels. 
Each frame shows the PSF radial profile and intensity map of one color channel. Star test measurements were made on an optical bench with an f/8 beam. The CCDs had 15m pixels. We used a pin hole with a size of 20m in the image plane. The radial profiles show the pixel values (crosses) together with a gaussian fitted to these points. The full width at half maximum (FWHM) of the gaussians is indicated. The angular scales given along the upper side of each diagram, at the gray scale images and for the FWHM values are calculated for the 2.2m f/8 RC telescope at the Calar Alto observatory for which BUSCA is designed. The intensity scaling for the pixel maps is logarithmic to enhance low intensity levels.

3.2 Filter wheels and filters

The dichroic beamsplitters preselect the wavelength bands only. Additional filters with well defined transmission curves have to be used to get as close as possible to standardized astronomical filter systems or to select the narrow band around emission and/or absorption lines respectively. Four filter wheel units have been integrated into the "cube" in such a way that the filters are immediately in front of the dewars (see Fig. 3). The compact size of the instrument on the one hand and the size of the filters (diam. = 110mm) on the other hand restricted the number of positions to four. In addition filters can easily be exchanged, only the dewar has to be taken away. This gives the instrument a high degree of flexibility. Filter wheels are driven by stepper motor-encoder combinations so that an independent position control is possible. Position changes take a few seconds.

It is planned to use BUSCA first of all for the study of individual stars and stellar systems. Consequently the Stömgren filter system will be available at first with filters having the full size (110mm). But filters of smaller diameter available at the observatory or supplied by observing groups could be mounted. In this case the unvignetted geometrical field of view is reduced to a size which is about 10mm smaller than the filter diameter (e.g. 40mm for a 2 inch filter., or about 7.5 arcmin).

3.3 The shutter

Precision photometry is only possible if exposure times are precisely known, and for the photometry of extended fields these exposure times have to be homogeneous across the whole field. With the growing size of monolithic CCDs and CCD mosaics in particular, the iris type or similar shutters – if available at all – produce unacceptable inhomogeneities which increase at shorter exposure times. In this situation groups building large format cameras have developed their own shutters. The ideal situation for BUSCA would be individual shutters for each CCD system. This would allow individual read outs depending on the light levels in the different color channels, which depend on the spectrum of the source, the overall filter transmission characteristic and the wavelength dependent QE of the detectors. But space limitations and restrictions set by the CCD controller have led us to the decision to place a single shutter in front of the first dichroic. At this position the shutter aperture has to be 110mm x 110mm. Our goal was to achieve homogeneous exposures down to exposure times of 0.1sec with remaining inhomogeneities of 1 percent or less. Shorter exposure times should be possible. Short exposure times are needed as calibration stars are usually very bright. With the increasing size of new telescopes this problem is even more severe. Another important aspect of short exposure times is that observers run often out of time if they try to get sky flat fields for calibration at dusk or dawn. With very short exposure times at their disposal this time can be extended.
 
 

Fig. 6  The shutter
The photograph on the left hand side shows the prototype of the shutter. Images and measurements presented in chap. 7 were made with this prototype. Its main elements are two carbon fiber blades moving along linear bearings, two stepper motors and tooth belts connecting motors and blades. The shutter aperture is 110mm x 110mm. The diagram on the right hand side shows the effective exposure times of a 10ms exposure as a function of the position across the shutter aperture. Measurements were made with a laser and a photodiode at equidistant positions across the shutter aperture (in the direction of the blade movement). A short exposure time of 10ms were chosen to make the photodiode output signal visible on an oscilloscope in a way that effective exposure times could be measured directly using the oscilloscope screen cursors. As the blades are always moved in exactly the same manner independent of the exposure time the inhomogeneity profile seen in the diagram would only be shifted along the time axis without changing its form.

We have built a slit type shutter as a prototype (see Fig. 6) which consists basically of two carbon fiber blades moving on linear ball bearings and which are driven by two stepper motors and tooth belts. We achieved a symmetric design without a preferred direction of movement. Consequently the shutter blades are moving from left to right and then from right to left for consecutive exposures. For practical reasons the two stepper motors are on the same side. This allowed to have a flat section inside the "cube". The entire shutter can be taken out for maintenance easily like a drawer. The stepper motors are combined with incremental encoders which serve as indicators in case of system failures (e.g. blocking of the shutter blades). Both blades are driven with identical velocity profiles, a prerequisite for homogeneous exposures. The velocity profile is always the same independent of the exposure time and can be set by software. The full movement of a blade takes about 0.2sec. For exposure times shorter than that (lower limit at about 1 msec) the two blades automatically form a moving slit. The precision of the exposure time and the homogeneity of the exposure is mainly limited by mechanical parameters (e.g. tooth belt tension).

The shutter performance measured in the laboratory (Fig. 6) shows that residual inhomogeneities in the inner 2/3 are well below 1 msec and they increase to up to 1.5 msec near the edges. This systematic trend can probably be compensated by fine tuning of the velocity profiles but these effects are smeared out anyway as the shutter is far from the focus. Dome flats of 0.1sec and 2sec show no systematic differences above the 1 percent level, corresponding to residual inhomogeneities of less than 1 msec. This is consistent with the laboratory results. The shutter is well suited for precision photometry down to very short exposure times of 0.2 sec and probably even 0.1sec, allowing to even take daytime sky flats.
 
 

BUSCA was designed to have for each color channel an independent CCD detector unit, all operated by a single CCD controller. The geometrical size of the detector area in the focal plane and the pixel sizes followed from considerations concerning the available CCD formats, the field of view given by the telescope where distortions without corrector lens are still acceptable, the sampling of the PSF, and the overall size of the system.

4.1 The CCD detectors

Since BUSCA is a simultaneous 4-channel photometer equipped with a separate CCD camera system for each of the 4 distinct wavelength bands, we have been looking for CCD solutions that cover the entire 6cm x 6cm FOV with high enough spatial resolution, and which have an UV sensitive CCD for the UV channel. Among different possibilities we have been studying the following choices: monolithic 4Kx4K, 15mm CCDs, monolithic 2Kx2K, 24mm CCDs (with a slightly smaller FOV of 5cm x 5cm), mosaics of 2Kx4K, 15mm CCDs, and mosaics of 2Kx2K, 15mm CCDs. Financial considerations led us finally to choose three thick, monolithic 4Kx4K, 15mm, LMFS CCD485 and one CCD485 thinned and coated by Mike Lesser for enhanced UV response. The three thick CCD485, originally designed for still-image photography¹⁰ show a quite low QE with its maximum of about 40% at 650nm. The thinned version will have an overall high QE response and will be optimized for the UV. As preliminary devices we use three thick, 2Kx2K, 15mm, Loral CCD442s, one of which was coated with METACHROMEÒ II for improved blue sensitivity, and one Loral 2Kx2K, 15mm PPtF CCD thinned and coated by Mike Lesser for the UV channel. Three thick LMFS 4k CCDs have been delivered and their implementation is under way.

4.2 CCD cooling

The cooling of CCDs at the Calar Alto observatory is based on the classical liquid nitrogen bath dewar, a system which we also used for all of our CCD cameras developed in Bonn so far (mostly 2Kx2K systems). The dewars were designed and built in Bonn and we decided to construct the BUSCA dewars as well. The major difference to our standard dewar is that they had to be large and should function independent of the orientation. As the dewars are sitting around the "cube" a dewar mounted on the western side would get into an upside down position if the telescope is looking eastward. We have chosen the most common and simple solution by limiting the length of the fill-in tube inside the liquid nitrogen so that only one half of the available tank volume can be filled. These dewars have been tested so far in the lab still without a detector. We got hold times of more than 24 hours which will be completely sufficient as at the observatory CCD dewars are routinely filled before and after each observing night.

CCD operating temperatures are from –60C to –100C. For the temperature control we use commercially available self calibrating PID controllers which can be remotely accessed over a RS485 bus. They are part of the instrument control unit.

Experience at all observatories shows that dewar hold times may be subject to sudden changes for a variety of reasons. With four dewars there is a non negligible probability that such problems may occur during an observing period. Therefore the dewar hold times are routinely monitored by a simple method which acts primarily as an alarm system. NTC resistors are used as temperature indicators for the evaporating nitrogen. They are connected to the ADC inputs of a microcontroller which is part of the instrument control unit (see below). NTC resistors are placed inside the four flexible hoses which guide the nitrogen gas from the dewars into the "cube" in order to avoid dew problems particularly on dewar windows. One additional sensor is mounted outside measuring the ambient temperature for reference. The sensors inside the hoses are close to the dewars so that temperature differences are several tens of degrees. (The nitrogen gas when entering the "cube" is at ambient temperature.) Experience shows that after the nitrogen in a dewar has completely evaporated the temperature difference drops to zero within a few minutes. The heat capacity of the copper nitrogen tanks is sufficiently large to keep the CCD at its set temperature for more than one hour. After a "dewar empty" alarm there is plenty of time to finish a running exposure and refill the dewars .

4.3. The CCD controller

Already during the initial phase of the project it was clear that there would be up to four CCD outputs for each of the four detector units. In order to make read-out times as short as possible a fast 16 channel system would have been the ideal solution. But due to a number of restrictions (observatory policy, time and budget limitations) we ended up with a four channel system. The CCD controller has been designed at the MPIA to be able to drive four CCDs (4 independent clocking units and CDS/ADC units). The BUSCA controller, assembled and tested in Bonn, is the first system where this option had to be realized.

The central unit of the controller is a DSP board where all major operations are performed (command communication, telemetry communication, data transmission, initialization of the parallel shifts). The read out and A/D conversion clocking sequence for a whole CCD line is produced by the clocking units. But the fact that the transmission of each single word of pixel data (or word quadruple for 4 CCDs) is organized by the DSP sets an upper limit to the read-out speed of about 50Ksamples/sec. The read-out time for a complete 4Kx4K CCD is then around 5 minutes. Fortunately the average seeing conditions at the observatory (around 1.5 arcsec) allow a 2x2 pixel binning to be used – corresponding to 0.35 arsec per binned pixel - so that read-out times in most cases are reduced to less than 100sec.

We have measured read-out noise values of about 6 electrons for our 2Kx2K LORAL devices at all four channels where the controller electronics contributes less than 1 count.
 
 

Apart from the operation of the CCD detectors a number of control and monitor functions have to be covered. All those functions are integrated in the instrument control unit (Fig. 7): Driving of the filter wheels, driving of the shutter blades, control of the CCD temperatures, monitoring the temperature of the evaporating nitrogen gas. When designing the instrument control unit it was not clear what kind of functions apart from the obvious ones would have to be added later– particularly monitor and metrology functions. But from earlier experience with "growing systems" leading – among other things - to a growing number of communication connections it was obvious that a flexibly expandable control system with only one single communication line to the control computer was highly desirable. The observatory requirement to have access to all control functions by typing simple ASCII commands at a terminal independent of the control computer made the "single communication line" solution even more favorable.
 

Fig. 7 Instrument control unit.
A half height 19” rack contains 8 microcontroller (mC) systems (single boards) based on the C515C and four commercially available temperature controllers for the CCDs. mC boards are all identical. Their individual function is set by the Eprom software. They are nodes on a CAN bus. A single RS232 line connects the unit with the control workstation.

Heterogeneous functionality among connected and freely accessible subunits is one major aspect of what is usually called a "distributed system". The type of connection between subunits plays the most important role as this is where the openness of the system is based on. Over the past few years serial field busses became a very common and very cheap solution to this problem. Subunits (bus nodes) are then often built around microcontrollers. We have selected the generic CAN field bus system. The main reasons for selecting the CAN bus are that cheap hardware and software components on the market make own developments of bus nodes feasible. We could build all our subunits around the same microcontroller type (Siemens C515C) which has an integrated CAN controller. Although the functionality of individual subunits is slightly heterogeneous one single hardware design has been found which meets all our needs. This of course has many advantages for production and maintenance. The required functions are:

1. Several stepper motors have to be driven and their positions have to be monitored by means of incremental encoders (mounted on the motor axes). The filter wheel units as well as the shutter have this stepper motor/encoder combination.
2. Access to the RS485 bus which connects the CCD temperature controllers.
3. The NTC sensors (inside the nitrogen exhaust hoses and at ambient temperature) have to be monitored.
4. Command communication between the instrument control and the control computer (workstation).

The data acquisition system of BUSCA consists of the MPIA controller, a controller interface box and an Alpha-AXP workstation running Digital UNIX. The data connection between the controller and the workstation is divided into a slow (4800 baud) serial line and a high speed (tested up to 2 MBytes/s), 16-bit wide, parallel line, which uses the Digital DRV11 standard implemented over a Logical DCI-1100 PCI-interface board at the workstation. This interface board uses FIFO buffers and 32-Bit DMAs for data transfers to and from the workstation. The serial line is used bi-directional to send commands to the DSP, the main signal processor of the controller, and to receive its answers, while the parallel line is used almost exclusively one-directional to receive the image data from the controller. The different software layers are the DSP program, the UNIX device driver for the Logical board, and a C-Library, with interface routines. For evaluation purposes an event-driven user interface was created under IDL (Interactive Data Language) to allow the initialization of the controller, the simultaneous data acquisition of up to 4 read-out channels, the display and storage of image data in FITS format and evaluation of primary CCD parameters (RON, CTE, full well).
 
 

BUSCA was tested during two observing runs (September, November 1998) at the 1m f/15 cassegrain telescope of the Hoher List observatory of the Univ. of Bonn. The observatory is located in the Eifel mountains close to Bonn. The instrument was equipped with four 2Kx2K CCD systems as described in sections 3.1 and 4.1. At the 1m telescope our 2Kx2K 15mm pixel CCDs give a field of view of 7 arcmin x 7 arcmin. First light was at September 11 but weather conditions did not permit systematic work during this first run. In the November campaign observations could be performed during several clear nights. They were mainly aimed at demonstrating the imaging capabilities, studying atmospheric transparency fluctuations and their influence on color indices and at the calibration of the Strömgren filter set. Filters of smaller diameter (2 inches), which are available at the observatory, could be used without vignetting the field covered by the 2Kx2K CCDs. These were mainly line filters (e.g. Ha, Hb, O[II], O[III]).

Fig. 8 shows four images of the crab nebula taken simultaneously. The nebula originates from a supernova explosion. The images seen through narrow band line filters (O[II], Ha, Hb), showing mainly filamentary structures, are dominated by emission from the hot ionized matter which was expelled by the supernova explosion. The underlying smoothly distributed continuum emission from the inner part of the nebula (mainly synchrotron emission) is seen best in the broadband NIR image. Images were aligned by rotation/translation. As expected no distortions of the four images relative to each other were found down to a limit of about 0.5 pixel over the whole field.
 

Fig. 8 The Crab Nebula seen with BUSCA
These images have been taken simultaneously in the four BUSCA channels (uv_B: O[II]; b_B: Hb; r_B Ha; nir_B: NIR). Narrow band interference filters were used for the emission lines and an RG780 low pass filter for the NIR band. Exposures were taken in a series of 6x 10min. CCD pixels were binned 2x2 (0.4 arcsec). The continuum emission contribution in the emission line images was not subtracted. 

During periods of varying atmospheric transmission we have taken series of exposures to study the spectral behavior of transmission changes. Transmission changes which are not gray would directly influence the determination of the intensity ratios between different color bands (i.e. color indices), even for simultaneous measurements. Atmospheric extinction in the visual is mainly due to scattering by particles of various sizes (from molecules to droplets and dust grains) with a wavelength dependence l^-a. The mixture of particle sizes in the atmosphere leads to an average spectral index of a@  1 and to a general reddening effect on spectral measurements² which – if stable under photometric conditions - can be corrected for. Short term extinction fluctuations are often found to be gray². They are most probably due to large particles (droplets, ice crystals, dust grains). The analysis of our "color of extinction" measurements is still in progress. Here we present preliminary results.
 

Fig. 9 The color of atmospheric extinction fluctuations.
a) – c) Effect of atmospheric extinction fluctuations on intensities mxxx and intensity ratios (mxxx - myyy) (i.e. color indices) for three stars. The intensity scale is logarithmic with arbitrary zero points. The three curves correspond to three Stömgren bands v,b,y. Indices xxx, yyy give the center of the corresponding wavelength bands in nm. Bandwidths are from 20nm to 40nm. Individual exposures were 5 minutes long and were started every 7 to 8 minutes. Intensity measurements of a given band are connected by a smooth line to avoid confusion. Average intensity ratios are marked by a straight line. Individual curves are offset with respect to each other by an arbitrary amount. The apparent magnitudes mV of the 3 stars are 9.3(1), 11.2(2), and 13.7(3). 
d) This diagram shows the same (m410-m470) values as in a) in comparison with intensity ratios calculated according to the usual sequential method: each number pair giving a (m410 - m470) value was taken from consecutive exposures. Scaling is as in a)-c).

Fig. 9 shows data of three stars in a field which was monitored over 80 minutes. 11 exposures, 5 minutes each, were taken. During this time the passage of thin clouds led to a significant drop in measured intensities (about a factor of 7, definitely non-photometric) which is apparent in all three Strömgren bands shown in Fig. 9. Color indices determined in the classical sequential manner (see Fig. 9d) would give nonsense results. The color indices derived from the simultaneous measurements (Fig. 9a-c) are largely constant with a scatter of about 2 percent. The scatter becomes larger for fainter stars. In the case of the brightest star (star 1) a weak systematic trend is apparent in the intensity ratios of the outer bands (410nm, 550nm). This is seemingly correlated with the direct measured intensities, suggesting that the extinction variation was not completely gray. The peak to peak variation of the intensity ratios is 8 percent. On the other hand, these systematic color index variations are towards the blue which is difficult to explain. In any case, if this is compared with the drop in intensity of a factor of 7 it means that the color index variation scales with the total intensity variation like 1/100, probably less. Therefore it can be concluded that, if transmission changes by thin clouds are only small (a few times ten percent, still non-photometric), this would have a negligible effect on simultaneously measured color indices.
 
 

With BUSCA the efficiency and quality of spatially resolved multicolor photometry will be significantly improved, due to simultaneity. First tests at the telescope have shown that, apart from the obvious gain in observing time, the precision of color index determination is considerably increased and largely independent from short term atmospheric extinction variations to the extent that reliable color index measurements will be possible even under non-photometric conditions.

The implementation of large format 4Kx4K CCDs is underway. This will be the last major step before a first observing run at the Calar Alto observatory. It is planned to release BUSCA as a common user instrument for the winter semester 1999/2000.
 
 

O. Cordes has helped with the observations and the preparation of the observational data presented in chapter 7. The CCD controller was designed at the Max Planck-Institute for Astronomy (Heidelberg/Germany). We kindly acknowledge support and advice by B. Grimm of the MPIA. The BUSCA project was realized with financial support by the Bundesminister für Bildung, Wissenschaft, Forschung und Technologie through grant 05 3BN114 (4).
 
 

1. E.H. Geyer and M. Hoffmann, "A double beam photoelectric photometer for astronomical applications", Astron. Astrophys., 38, pp. 359-362 , 1975.
2. Chr. Sterken and J. Manfroid, Astronomical Photometry A Guide, Kluwer Academic Publishers, Dordrecht, 1992.
3. H. Barwig, R. Schoembs and C. Buckenmayer, "A multichannel multicolor photometer for high time resolution", Astron. Astrophys., 175, pp. 327 - 344, 1987.
4. S. Marshall, "The CCD array camera for the macho project", in .Astronomy from wide-field imaging, H.T. Mac Gillivray et al., eds., pp. 67 – 71, Kluwer Academic Publishers, Dordrecht, 1994.
5. M. Doi, H. Furusawa, F. Nakata, S. Okamura, M. Sekiguchi, K. Shimasaka and N. Takeyama, "UT 15-color dichroic-mirror camera and future prospects", Proc. SPIE, 3355, pp. 646-657, 1998.
6. K. Meisenheimer, S. Beckwith, R. Fockenbrock, J. Fried, H. Hippelein, U. Hopp, Ch. Leinert, H.-J. Röser, E. Thommes and Ch. Wolf, "The Calar Alto deep imaging survey", in The early universe with the VLT, J. Bergeron, ed. , pp. 165-172, Springer, Berlin, 1997.
7. A.J.P. Theuwissen, Solid-State Imaging with Charge-Coupled Devices, Kluwer Academic Publishers, Dordrecht, 1995.
8. M.I. Andersen and A.N. Sorensen, "An interferometric method for measurement of the detector MTF", in Optical Detectors for Astronomy, J.W. Beletic and P. Amico, eds., Astrophysics and Space Sience Library, 228, pp.187-190, Kluwer Academic Publishers, Dordrecht, 1998.
9. P.R. Jordan, A.P. Oates, S.M. Tulloch, D.J. Ives and X. Gao, "New generation CCDs- application and performance results", Proc. SPIE, 3355, pp. 646-657, 1998
10. R. Bredthauer, "The state of detector technology at Lockheed Martin Fairchild Systems", in Optical Detectors for Astronomy, J.W. Beletic and P. Amico, eds., Astrophysics and Space Sience Library, 228, pp.13-18, Kluwer Academic Publishers, Dordrecht, 1998.