IL NUOVO CIMENTO
VOL. 18 C, N. 3
Maggio-Giugno 1995
The Solar-Image Acquisition System at Tor Vergata University. F. BERRILLI, S. CANTARANOand A. EGIDI Department of Physics, University of ~Tor Vergata,, V.le della Ricerca Scientifica 1, 1-00133 Roma, Italy (ricevuto il 2 Gennaio 1995; approvato il 31 Matzo 1995)
Summary. -- We describe an image acquisition system realized as a part of an apparatus built in collaboration with the Arcetri Astrophysical Observatory in Florence designed to record high-spectral-resolution solar images in the visible part of the spectrum. The system is based on a 512 • 512 Thomson CCD type THX31159 and on a 486 CPU personal computer running under MS-DOS. The electronics for driving the sensor and for the amplification and conditioning of the video signal has been designed and built in the laboratory while the signal A/D conversion and image presentation is performed using commercial boards. PACS 96.60 - Solar physics. PACS 95.55,Mc - Astrophysical (including spectroscopic) instrumentation.
1. -
Introduction.
The Astrophysics and Cosmic Physics Group of Tor Vergata University and the Arcetri Astrophysical Observatory are jointly performing a series of high-spectralresolution observations of solar absorption lines. The primary observational goal is to measure profiles of the spectral lines as a function of time over the entire disk. The scientific objectives which are now pursued are the study of line oscillations and velocity fields. The instrumentation, installed at the Arcetri Solar Tower, consists of a tunable Universal Birefringent Filter (UBF) coupled with a Fabry-Perot (FP) Interferometer, having an overall bandwidth of 2.5 pro, followed by an image acquisition system using a 512 x 512 CCD sensor controlled by a PC-based electronic apparatus. In this paper we describe the acquisition system, while we refer the reader to [1] for a description of the optical and interferometer system. In particular, sect. 2 describes the CCD Camera and the driving electronics, sect. 3 deals with the PC-system, sect. 4 presents the calibration measurements, and finally sect. 5 presents conclusions and future developments. Since the significance of oscillation studies depends obviously on the time rate at which images can be acquired, care has been taken to reduce the acquisition time to a minimum compatible with a photometric error of the order of one part in a thousand. The apparatus described here is designed to acquire about ten images per minute, which allows, for instance, the study of 5-minute oscillations of a spectral line sampled at least at five different wavelengths. 269
270
F. BERRILLI, S.
CANTARANO
and A. EGIDI
2. - The CCD camera and electronics. The sensor used in the camera is the Thomson THX31159 CCD, a square device composed of 512 x 512 pixels, 19 x 19 ,~m2 each, which is front illuminated, having a photosensitive area of 9.7 x 9.7 mm ~. The device is built with n-buried channel technology for high charge transfer efficiency at low signal level. The sensitivity range goes from 400 to ll00nm with a peak quantum efficiency of 0.4 at 700nm (nominal). The typical dark current is 5 e-/pixel/s at - 45 ~ Charge integration and row transfer are controlled by fourphase drive clocks obtaining the maximum of charge storage capability in the photosite; the serial register is read out by a two-phase drive dock. Two outputs are provided through two on-chip amplifiers with different gains and bandwidths. In order to reduce the dark current, the sensor is cooled by means of a two-stage solid-state Peltier device, which in turn is water cooled; the temperature of the sensor is held at about 50~ below the water temperatm~e. To avoid water condensation, the detector is kept under vacuum in an aluminum container having a window consisting of a hot-mirror IR filter. The video signal preamplifier, the phase drivers and a termocouple amplifier are contained in the camera. The termocouple measures the temperature difference of the CCD chip with respect to the cooling water. The control circuitry, entirely developed in the University laboratory, is realized on Eurocard boards mounted in a standard 5" rack container and is composed of the following elements (fig. 1): i) The Signal Conditioner, which measures the charge deposited on each pixel by means of a double correlated sampling. This technique consists in sampling the video signal at the end of the output gate reset period and again soon after the pixel charge has been transferred on the output gate and taking the difference of the two signals. This method allows to achieve a higher speed with respect to the dual slope integrator scheme. The readout time is 8 ~s/pixel, corresponding to a complete CCD image readout in about 2 seconds. The gain can be varied by means of jumpers, to optimize to the various working conditions. ii) The CCD power supply, generating adjustable voltage levels for the CCD clock drivers. iii) The CCD logic circuitry, which generates the clock pulses necessary to operate the device. It can be driven either from an internal clock source or by the acquisition computer. iv) The Shutter control, which operates a Uniblitz 225L shutter. v) The Feltier power supply and monitoring. vi) The UBF interface, which handles the signals exchanged with the computer controlling the operation of the spectrograph. 3. - The PC-based system. The system is controlled by an AST 486/25 (25 MHz) Personal Computer running under MS-DOS and equipped with 8 Mbyte of RAM. With this memory the system can store up to ten images consisting of 512 x 512 words while running the acquisition program.
271
THE SOLAR-IMAGE ACQUISITION SYSTEM AT TOR VERGATA UNIVERSITY
f I
_ .
c
J
a
I
video signal conditioner
D C
CCD power supply
j "-I
Peltier power and c o n t r o l
I
shutter control timing
I
CCD clock generator
I
uoF
i
j
host computer
m e r a
interface
JsPectr~176176 r]
Arcetri computer
I
Fig. 1.- Block diagram of the CCD Controller architecture. Since uninterrupted observational runs of several hours are required at an acquisition rate of the order of 5 Mbyte/minute, the computer is connected to an external mass storage unit consisting of two 1.2 Gbyte Micropolis hard disks (for fast on-line image storage) and an Exabyte 8200 tape unit using 8 ram-2.2 Gbyte tape cassettes (for off-line storage). These mass storage devices are supported by a SCSI Future Domain TMC-1660 host adapter board. At the mentioned acquisition rate of about 5 Mbyte/minute (corresponding to 10 images per minute) the storage capacity allows about 6.5 hours of acquisition. The digital Input/Output interfacing with the CCD controller, the digitizing of the video signal, and the high-resolution video display are performed by appropriate boards plugged on the PC-bus (fig. 2). The Digital I / O board is a Burr-Brown PCI-20087W-1. This card features 40 digital I / O channels grouped into five eight-bit ports. The I / O lines are used for command and information exchange with the CCD controller (i.e. CCD operations, shutter and cooler operations) and for the exchange of handshaking signals with the Arcetri spectrometer controller. The A/D converter is an RTI-860 Analog Devices high-speed data acquisition board. This card provides 16 single-ended analog input channels multiplexed into a 12-bit Analog-to-Digital Converter providing a resolution of 2.44mV in the +5V used range. The typical analog-to-digital conversion time is 3 ps. An external pacer source, generated by the CCD logic board, synchronizes the A/D conversions
F. BERRILLI, S. CANTARANO
272
I Io ol
PI - 1024 video board AST 486/25 8 MB RAM 200 MB HD
A. EGID1
I
hi-resolution monitor
VGA monitor
and
2.2 GB tape
TMC - 1660 SCSI controller 9~ RTI - 860 A/D converter 9~
CCD signal CCD temp. photodiode
PCI - 20087 W- 1 I/O board CCD controller Fig. 2. - The computer system. bypassing the internal clock of the RTI-860 card. A whole 512 x 512 image is stored in the on-board buffer and then transferred to the computer's RAM. Two more channels are used to acquire the temperature difference across the Peltier cooler of the CCD device and a photometric signal from a photodiode. A PIP-1024 Matrox board provides a video display with a resolution of 512 x 512 pixels on a Mitsubishi HF3400 high-resolution monitor. The program controlling the whole measurement cycle has been written in C-language making use of the board specific software libraries of the RTI-860 A / D converter, of the PCI-20087W-1 I / O board and of the PIP-1024 Matrox video card. A special video library [2] is used for monitoring and passing parameters to the program interactively through the use of menus and windows. Application programs for data reduction, analysis and display have also been developed, using Lahey EM/32 FORTRAN language. 4.
-
Calibrations.
In order to optimize the performance of the device and to reduce the electronic noise, intensive testing was carried out by varying several of the electronic parameters, as voltage levels for the CCD driver clocks, reset pulse and sampling pulse timing. Once the correct set of the parameters was defined, the system was characterized by measuring i) the linearity of its response, ii) the transfer and the conversion factor [3] and iii) the system noise. i)
Linearity.
To check the linearity of the system, the signal is recorded in a series of increasing time intervals under constant illumination; until the charge collected is less than about 30 per cent of the nominal maximum collecting capability, the linear correlation
THE SOLAR-IMAGE ACQUISITION SYSTEM AT TOR VERGATA UNIVERSITY
273
coefficient is greater than 0.99995. Deviations from linearity are found for a number of collected electrons per pixel greater than 100 000. This behaviour appears to be due to an intrinsic property of the CCD and not to an incorrect drive of the device. ii) Transfer and conversion factor. The ratio g between the number of photoelectrons collected on a pixel and the output voltage expressed in levels of the A / D converter (ADU) characterizes the behaviour of a CCD system [3]. The coefficient g and its inverse G, the so-called ,transfer factor,, were determined by applying the photon transfer curve technique described by Janesic et al. [4, 5]. This technique is described in the following. ~,Flat, images at different levels of illumination are acquired. The mean and the standard deviation of the signals of each of them are computed after removing the voltage offset, introduced by the electronics and by the dark current, and the spatial non-uniformity of the illumination and of the detector's response. By plotting on a logarithmic scale the standard deviation vs. the signal, both measured in ADU, a curve is obtained in which we can identify two asymptotes of different slopes. The one at high values of the signal represents a signal dominated by the photon noise with a slope equal to 1 / 2 (due to the statistical nature of photon emission). The other one represents a signal dominated by overall noise and has a zero slope. The intercept of the first asymptote with the abscissa gives the value of the coefficient g from which the transfer factor G is derived, while the level of the horizontal line gives the value of the overall noise. iii) Overall noise. The overall noise consists of the read-out noise, the CCD thermal noise, and the noise of the amplification chain. Of these, the read-out noise is usually the largest. The CCD thermal noise is due to the fluctuations associated with the dark current produced by the hole-electron pairs generated by thermal vibrations of the silicon lattice. The dark current, and the related thermal noise, is determined by measuring the collected charge in the non-illuminated sensor as a function of integration time. The noise of the amplification chain is usually less than the CCD thermal noise. The following results, G-factor 0.95 and 2.1 ( ~ V / e - ) and overall noise 61 and 70 (e-), come from the tests performed on two units of the system. The photon transfer curves are shown in fig. 3. While the differences in the overall noise are inherent to the two CCD devices, the different conversion factors are due to the fact that in the first system the low-gain on-chip preamplifier was used. A similar value of dark current has been found for the two systems: 50 (e-/spixel) at a temperature T = - 4 0 ~ 5. -
Conclusions.
The PC-based acquisition system presented is a compromise between astronomical CCD systems for nocturnal use, which require lower noise but allow a longer time for read-out, and video-rate CCD systems, which are faster but have fixed integration time, higher noise and usually a smaller resolution due to the use of 8-bit AD converters. The system has been extensively tested in the Arcetri Astrophysical Observatory
274
F. BERRILLI, S. CANTARANOand n. EGIDI
1.2
-r~j--r-i
~
i
,
i
~ i
~ i
~ i
~ I
,
1.0
h~ 0.8
0.6 1.8
2.2
2.6 3.0 logsignal(ADU)
3.4
Fig. 3 . - P h o t o n ~ a n s f e r c u ~ e o f t h e ~oCCDs(mCCD #1, ACCD #2). to acquire high-spectral-resolution images of the full disk of the Sun obtained through a Universal Birefringent Filter and a Fabry-Perot interferometer. The data acquired, altough preliminary and mainly intended to assess the performance of the experimental apparatus, have provided information on solar oscillations[6] and detailed profiles of absorption lines over the entire disk of the Sun. The system can be used also for high-spatial-resolution imaging. In this case a second system is necessary to acquire a white-light image of the field for the purpose of eliminating the deformations of the image due to atmospheric turbulence (,,destretching~0. The system in now being upgraded to a new version, to be installed in the future solar telescope THEMIS in the Tenerife island, in which the CCD camera electronics is connected to the computer by an optical link and the camera electronics will be easily adjustable to CCD sensors of any size up to 1024 x 1024 pixels. The authors gratefully acknowledge the staff of Arcetri Astrophysical Observatory in Florence (Italy) for their collaboration during the testing runs and Dr. P. Cerulli-Irelli of the CNR-IFSI for his help in the development of the acquisition programs. REFERENCES [1] CAVALLINIF., CEPPATELLI G., RIGHINIG. et al, Nuovo Cimento C, 15 (1993) 509. [2] CERULLI-IRELLIP., CNR-IFSI Internal Note (1992). [3] MCLEAN I. S., Electronic and Computer-aided Astronomy (Ellis Horwood Limited Publishers, Chichester) 1989. [4] JANESIKJ. R., ELLIOTT., COLLINSS., BLOUKEM. M. and FREEMANJ. W., Opt. Eng., 26 (1987) 692. [5] JANESIKJ. R., KLAASENK. P. and ELLIOTT., Opt. Eng., 26 (1987) 972. [6] BERRILLIF., CACCINF., CANTARANOS. et al., Mem S.A.It., 64 (1993) 549.