The charge-coupled device (CCD) dominates an ever-increasing variety of scientific imaging and spectroscopy applications. Recent experience indicates, however, that the full potential of CCD performance lies well beyond that which is realized in currently available devices. Test data suggest that major improvements are feasible in spectral response, charge collection, charge transfer, and readout noise. These properties, their measurement in existing CCDs, and their potential for future improvement are discussed in this paper.
In this paper we report on new, fully manufacturable, high sensitivity frame-transfer-type charge coupled device (CCD) image sensors developed for high performance television applications including low light level surveillance and color broadcast cameras. A novel feature of these charge coupled device (CCD) imagers is a thinned configuration which allows illumination from the backside of the device, providing very high quantum efficiency in excess of 60% over the entire visible spectrum for the whole imager area with anti-blooming drains. This high quantum efficiency, together with the very low noise charge coupled device (CCD) read-out capability of less than 35 electrons root mean square (rms) per pixel provides outstanding low-light-level camera performance.
This paper presents a solid state image sensor intended for use in 625 lines color TV cameras. The "Line Transfer" organization allowed, using 3 p, design rules over a 6,6 x 8,8 mm format, the realization of an imager whose main performances are : 576 lines by 462 pel, minimum Scene illumination : 30 lux and good protection against smearing and blooming The line transfer organization uses conductive buses to transfer photoelectric information in the vertical direction. The interesting feature of this organization is its ability to separate the signal charges from the spurious ones through a double reading of the bus voltage. The principle of operation and specific performances are first described and expressed in terms of vertical transfer inefficiency. The mechanism of transfer inefficiency is analyzed. Its application to the bus's reading of a line transfer imager is detailed and the main performances are deduced.
New concepts for operating charge coupled devices (CCD's) in a multiple frame transfer mode which allow very high speed imaging of transient phenomena are described. These methods coupled with the inherent advantages of slow scan readout, including high dynamic range, low noise and excellent linearity, provide unique, new observational capabilities. Appropriate masking techniques allow a large portion of the COD to be employed as a high speed image buffer. Applications include spatial imaging of rapidly moving objects, observations of transient spectral chemical phenomena and the study of rapid physiological motion.
A liquid crystal attenuator (LCA) operated as a variable neutral density filter has been attached to a charge-coupled device (CCD) imager to extend the dynamic range of a solid-state TV camera by an order of magnitude. Many applications are best served by a camera with a dynamic range of several thousand. For example, outside security systems must operate unattended with "dawn-to-dusk" lighting conditions. Although this can be achieved with available auto-iris lens assemblies, more elegant solutions which provide the small size, low power, high reliability advantages of solid state technology are now available. This paper will describe one such unique way of achieving these dynamic ranges using standard optics by making the CCD imager's glass cover a controllable neutral density filter. The liquid crystal attenuator's structure and theoretical properties for this application will be described along with measured transmittance. A small integrated TV camera which utilizes a "virtual-phase" CCD sensor coupled to a LCA will be described and test results for a number of the camera's optical and electrical parameters will be given. These include the following camera parameters: dynamic range, Modulation Transfer Function (MTF), spectral response, and uniformity. Also described will be circuitry which senses the ambient scene illuminance and automatically provides feedback signals to appropriately adjust the transmittance of the LCA. Finally, image photographs using this camera, under various scene illuminations, will be shown.
Since the introduction of the CID (Charge Injection Device) sensor by General Electric, there has been steady progress in reducing noise and artifacts as well as improvements in effective video bandwidth. New sensor technology, coupled with camera design techniques which use the advantages of the CID structure, have substantially reduced noise and made possible increased video bandwidth. Cameras with no apparent fixed pattern noise (spatial noise) have been demonstrated -without requiring specially selected sensors. Also, random ("white" or temporal) noise has been reduced to about .1% of saturation current (at a bandwidth of 4.2 MHz). Additionally, a technique has been demonstrated which extends camera bandwidth to 4.2 MHz (± 1 dB) with current technology sensors.
The design and development of a monolithic 1024 x 1024 pixel solid state imager will be discussed. The imager is a silicon MOS device with dual CCD readout registers. The design and processing of the imager will be reviewed and the camera electronics will be described. Performanceof the imager at this point in its development will also be discussed.
A line-scan image sensor with color sensitivity has been developed. The sensor utilizes a green-white-yellow-cyan sequence of pixels so as to provide one sample of luminance data for every two pixels and one sample-set of chrominance data for every four pixels. The color filters are formed directly on the silicon device. Good performance has been demonstrated. This device is expected to find use in industrial inspection cameras where color is important.
This paper presents the design method to achieve ultra-high resolution linear imagers. This method utilizes advanced design rules and novel staggered bilinear photo sensor arrays with quadrilinear shift registers. Design constraint in the detector arrays and shift registers are analyzed. Imager architecture to achieve ultra-high resolution is presented. The characteristics of MTF, aliasing, speed, transfer efficiency and fine photolithography requirements associated with this architecture are also discussed. A CCD imager with advanced 1.5 um minimum feature size was fabricated. It is intended as a test vehicle for the next generation small sampling pitch ultra-high resolution CCD imager. Standard double-poly, two-phase shift registers were fabricated at an 8 um pitch using the advanced design rules. A special process step that blocked the source-drain implant from the shift register area was invented. This guaranteed excellent performance of the shift registers regardless of the small poly overlaps. A charge transfer efficiency of better than 0.99995 and maximum transfer speed of 8 MHz were achieved. The imager showed excellent performance. The dark current was less than 0.2 mV/ms, saturation 250 mV, adjacent photoresponse non-uniformity ± 4% and responsivity 0.7 V/ μJ/cm2 for the 8 μm x 6 μm photosensor size. The MTF was 0.6 at 62.5 cycles/mm. These results confirm the feasibility of the next generation ultra-high resolution CCD imagers.
Multi-anode microchannel array detector systems with formats as large as 256 x 1024 pixels are currently under evaluation in the laboratory. Preliminary performance data for sealed ultraviolet and visible-light detector tubes have shown that the detector systems have unique characteristics which make them complementary to photoconductive array detectors, such as CCDs, and superior to alternative pulse-counting detector systems employing high-gain MCPs.
Charge-coupled-device (CCD) arrays are combined with an image intensifier tube to create an x-ray imager. The x-ray image intensifier tube design uses a proximity focus arrangement with four 32 x 32 CCD arrays.
A new version of an intensified charge-coupled device (ICCD) system, which is based on the incorporation of a CsI mesh photocathode and a thinned backside-illuminated CCD into a Space Telescope design Digicon tube, is being developed for future use in space-based observations in the Extreme Ultraviolet waveband. In this paper, we report the results of computer simulations involving the effects of the non-uniform electrostatic field near the mesh photocathode on imaging properties (especially, the spread function and modulation transfer function) of the detector system. The results are of importance in the design of the front end of the tube and the selection of optimum tube operating conditions. The overall resolution characteristics of the tube, including the effects of the mesh, the electron optics of the region between the mesh photocathode and the CCD target, and the structure of the CCD itself, are estimated.
The design and performance of a machine vision camera fabricated in thick film microcircuitry is described. The camera is manufactured with either a 128x128 or 256x256 pixel image sensor mounted onto a hybrid microcircuit. The 256x256 pixel front hybrid is constructed using multiple stratified layers of conductor on dielectric. Resistors are screened and trimmed on top of dielectric, a state-of-the-art procedure. Both 128x128 and 256x256 cameras use a common rear substrate which incorporates the clocking circuitry for the camera. The hybrids are formed into a "sandwich" around an aluminum heat sink and differential I/O lines are taken to the camera interface connector by means of a flexible printed circuit. Video pixel rates to 8 MHz are obtainable from both cameras which allow frame rates to 380 frames per second with the 128x128 version. Additional camera features permit the user to activate a real time frame reset as well as take external control of the row clock for even higher speed image processing.
This paper presents the development of an electro-optic sensor used in the assembly of a 130-mm long focal plane array. The sensor is a buttable 2048 x 96 element TDI (Time Delay and Integrate) CCD imager. The design, architecture and processing of this optical imager are described, as are particular problems associated with buttability and assembly requirements. The tap structures used for high frequency data acquisition are briefly discussed. The feasibility of this high-speed system for imaging of large land areas with high resolution and sensitivity, even under low optical contrast, is demonstrated. The performance specification data is presented.
A brief historical progression in infrared imaging from single element detectors to the state-of-the-art two-dimensional hybrid arrays is reviewed. A demand for higher device performance has revealed many of the disadvantages that are inherent with two-dimensional imaging array architecture. The advent of three-dimensional architectures is traced according to the evolution of these concepts. Several configurations are described and con-trasted. The future for three-dimensional imaging arrays is assessed based upon increased performance potential and the technical difficulties that must be overcome.
As a part of the NASA Imaging Spectrometer program, the Jet Propulsion Laboratory (JPL) has contracted with the Rockwell International Science Center to develop short-wavelength mercury cadmium telluride (HgCdTe) hybrid focal plane arrays. The imaging spectrometer concept requires high-performance array detectors which can be mosaicked into large focal planes. In the case of the Shuttle Imaging Spectrometer (SIS), a 64 x 64-element array with a cutoff wavelength of 2.5 μm is required. Prototype 32 x 32-element detectors with cutoff wavelengths of 4.8 μm and 2.5 μm have been tested by JPL. This testing has established confidence that the performance requirements for SIS can be achieved. The results have shown noise levels of 4000-6000 electrons per pixel for the 4.8 μm devices, dropping to 2000-4000 electrons per pixel for the 2.5 μm detectors. Pixel yields vary from 93% to 99%. The response linearity is exceptional for the 2.5 μm devices with a nonlinearity of about 2%. Quantum efficiencies as high as 60% have been measured. This paper will describe the results of the testing to date, including the performance of one of the 32 x 32-element arrays in the Airborne Imaging Spectrometer (AIS), a first-generation instrument which has produced extraordinary results in the identification and mapping of vegetation and geologic materials.
The design and performance of a standard television compatible, portable infrared charge-coupled device camera system is described. The sensor utilized is a 64 x 128 element Schottky barrier IR-CCD focal plane array cooled to cryogenic temperatures (77°K). The camera electronics generates the CCD clock waveforms necessary to drive the imager and contains all the video processing circuitry required to produce high quality IR imagery. The main feature of the video processing circuitry is a single-field subtraction circuit, using a field store with 8-bits of resolution, which suppresses fixed pattern noise (FPN), improves signal to noise (S/N), and decreases the minimum resolvable temperature. The RS-170 video signal output format enables direct interface to any standard video recording, processing or analysis equipment. A liquid nitrogen (dewar) cooling system and 12 volt D.C. power requirement permit portable applications of the system. Unlike scanning thermal imaging systems currently available, this prototype design is the first to incorporate a staring infrared focal plane array suitable for commercial applications. This camera demonstrates the high performance, design efficiency, and cost-effectiveness of a monolithic silicon Schottky barrier area sensor approach in an industrial infrared imaging system.
Results are presented on the performance of Schottky-barrier focal plane arrays operating in selected spectral bands. The devices measured were monolithic PtSi Schottky diode arrays, 64 X 128 elements, with buried channel CCD readout. The arrays were fab-ricated by RCA Laboratories, Princeton, New Jersey, under a contract from RADC/ESE for the development of high density focal planes. The spectral bands of interest were chosen in an attempt to determine the optimum filtering (spectral) for operation in day/night conditions with minimum interference from sun light or high clutter environments. The measurements were performed in both working thermal imaging camera systems and in laboratory dewars modified to accomodate filters for the desired spectral bands. Data are presented on the transfer characteristics and in-band thermal responses as well as the quantum yield of the devices. Conclusions about operation in the spectral bands chosen and comparison with broad band data are discussed. Examples of the high quality imagery obtainable with these devices are shown.
This paper describes a linear IRCCD focal plane consisting of a row of infrared sensing Schottky barrier photodiodes. These diodes directly "stare" at the infrared scene and staretime is automatically controlled by electronic circuitry without the need of shuttering or aperturing to obtain operation over the entire dynamic range regardless of the environment.
A unique method of fabricating silicon lenticular lens arrays has been developed to increase the effective fill-factor ratio of infrared focal plane arrays. Results on 32 x 63 element monolithic Schottky barrier (IR-CCD) devices have demonstrated greater than a factor of two improvement in device photoresponse. These miniature lenses are repeated in a uni-directional pattern to increase the optical fill-factor (in one dimension) to effectively 100%. Large arrays of these lenses have been fabricated on three inch silicon wafers, using a single photolithography step and unique three-step etching process. A novel procedure for lens array to device mounting and alignment have been developed to maximize lens efficiency. Near optimum performance was achieved and has been demonstrated through the application of lenticular arrays to Schottky barrier IR-CCD two dimensional focal plane arrays.
A new method has been developed to measure the modulation transfer function (MTF) of an array out to the Nyquist frequency without high-quality optical or mechanical components, without precision alignment, and with only one moving part. Test results for an infrared staring array of PtSi Schottky barrier construction show that this technique is a viable MTF measurement approach in the 3 to 5 pm spectral regions.
A low cost focal plane array test set was developed using a small micro-computer system controller. By careful selection of test types, hardware/software programability trade-offs and good software design the set performance has compared well to much larger test systems. An additional benefit of small size was noted--system noise levels were very low, enabling acquisition of high quality low level signal and noise data.
With the development of linear imaging arrays have come many applications for them. In order to use them, however, their performance has had to be experimentally determined, in detail. In light of potentially, highly sophisticated applications for current state-of-the-art linear imaging arrays, specific parameters for a selected set of linear imaging arrays were measured. The measured parameters were those usually never, if ever at all, presented in open literature from the manufacturers of these arrays or even in some cases not even known by the manufacturer. The specific parameters measured in order to characterize these arrays included crosstalk, total energy variation across an element, peak responsivity variation across individual elements and the full array, precise individual element size deter-mination and precise straightness measurements of the array's elements. The experimental setup utilized a HeNe laser source (632.8 nm) with precision optics to deliver a focused spot size onto the array measured in microns, a commercially available interferometric dis-placement measurement system to measure the position of the array relative to the focused spot and a computer to assist in the data taking and reduction. Two specific applications that immediately benefited from the measurements on these arrays were a precision star tracker and a precision angle sensor,' both under some phase of development at Lockheed Missiles & Space Company.
A computer-controlled CCD test facility has been developed based upon the Hewlett Packard HP9836C desktop computer. This paper describes the system and its operating software. Details of the control of the CCD operating parameters and some test procedures are described.
There is a symbiotic relationship between infrared imaging arrays and a new type of blackbody simulator whose design is based on the principle that the projected solid angle of the aperture is constant when viewed from all points of the cavity wall surface.
An advanced focal plane for remote earth resource sensing is under development for the Multispectral Linear Array (MLA) application. This paper will present a design description, the test methodologies employed, and recent test results.
The development of a CCD-based parallel analog processor is described. The singleinstruction,
multiple-data (SIMD) architecture allows substantial throughput improvements
when compared to conventional serial image processing hardware. The heart of the concept
is a single-chip array of analog processing elements interconnected with a two-dimensional
CCD shift register network. The same analog operation is performed on all cells
simultaneously. The device shows great promise in medium resolution (100 x 100) applications
(such as guided weapons and robotics) where it can be directly interfaced with a
staring focal plane through bump interconnections.
The Advanced Solidstate Array Spectroradiometer (ASAS) is a multispectral pushbroom scanner with 32 channels extending from 400 to 850 nm. It is built around a 32 by 512 element charge injection device (CID) array with enhanced sensitivity in the blue. It has twelve-bit output with variable gain and offset in the pre-amp and a framing rate which can be varied from 3 to 64 frames per second. The CID detector and the camera head electronics were designed and built by General Electric. The NASA Johnson Space Center (JSC) designed and built the data storage processor electronics. The optics were taken from an older system built by TRW for NASA. Contracting and coordination for the project was done under a Naval Ocean Systems Center (NOSC) program for ocean remote sensing, The completed system has been flown, tested, and calibrated and has been undergoing noise reduction analysis and modification at the NASA Goddard Space Flight Center. Details on the design, fabrication, and functioning of the CID array, the camera module electronics, and the data processor are given as well as an analysis of the scanner's performance and noise characteristics.
A star mapper using linear 2048 element CCD array is being developed for providing high accuracy post-facto attitude intormation for the Indian Remote Sensing Satellite (IRS-1A) Scheduled to be launched by the end of 1986. The sensor employs the orbital motion of the sun-synchronous satellite to map the sky at the orbital rate and generates coordinate intormation of the stars using an 1802 based processor system. A tour element 75mm aperature catadioptric system is used for sensing (stars upto fifth magnitude. The CCD is cooled by means of thermoelectric cooler to -20oC. Details of the sensor design are presented in this paper.
In early March 1986 the first Astro** mission will be launched to observe Halley's comet and a variety of other astronomical targets. Operating from the Shuttle bay, this payload consists of three large ultraviolet telescopes and a smaller wide field camera. An important part of the payload will be the first of a new generation of star trackers using CCDs as the star sensing element. The ASTROS tracker provides extremely precise measurements of star image coordinates as inputs to the Image Motion Compensation (IMC) system used to stabilize the science instrument focal planes. These coordinates, which reflect true star image motion to an accuracy of 0.2 arcsec, are required over a field of view of 2.2 x 2.5 degrees. This paper describes the design and application of ASTROS, with emphasis on performance test results acquired with a prototype system. Algorithms and software will be described in a later paper. Performance tests on real and simulated stars have consistently demonstrated 1/100 pixel accuracy and a noise equivalent angle of 1/300 pixel. These performance levels provide dramatic improvements, both in tracking accuracy and stability, relative to image dissector designs with comparable fields-of-view. These improvements, combined with other advantages inherent in a CCD-based approach are expected to lead to widespread application of this technology on future missions.
The Retroreflector Field Tracker (RFT) is an electro-optical position-measuring instrument which is part of the Dynamic Augmentation Experiment (DAE) to be used on the Solar Array Experiment (SAE), a NASA Marshall Space Flight Center (MSFC) experimental shuttle payload. The tracker measures and outputs the position of 23 reflective targets placed on a 32-meter solar array to provide data for determination of the dynamics of the lightweight structure. As such, the RFT is a noncontact optical position sensor that can be used in applications such as large space structure alignment, rendezvous and docking, and surface figure control of large antennas. The basic sensor operation, tracking logic, and position algorithms are extensions of concepts developed for star tracking. The solid-state sensor uses a 256 x 256 pixel charge injection device (CID) detector; the processor electronics employ three Z-80 microprocessors. The RFT includes a pulsed laser diode illuminator to project light onto retroreflective tape targets on the solar array. The RFT measures position to within *3 millimeters for all targets on the 32-meter array, resulting in an angular resolution of 19 arc seconds (1 sigma). The data output consists of target number, boresight angles, and orbiter coordinates. The data output rate is 2 hertz. The RFT is scheduled to be flown as part of the OAST-1 payload on Space Shuttle flight 41-D in Summer 1984.
Digital speckle pattern interferometry (DSPI) is a variation of electronic speckle pattern interferometry (ESPI). Both methods use the same basic techniques for nondestructive testing, but DSPI processes speckle pattern data digitally in a computer instead of using analog electronics to enhance fringe contrast. Both testing methods are an outgrowth of holography and speckle interferometry. The system described here uses a Reticon 100x100 diode array camera with an integration time of 5 ms, instead of a television camera, coupled to an HP-9836C computer. The use of digital methods provides flexibility in measurement and processing. Results of measuring both static and dynamic object movement show high-contrast fringes with electronic noise at 1/500 of the dynamic range. A new technique for testing vibrating objects has been developed that significantly in-creases fringe visibility. It involves subtracting a reference frame containing only self-interference terms and no cross-interference term from a time-averaged data frame of the object vibration. This reference frame is created by vibrating a reference mirror at a high amplitude while the object is at rest. Using software written in assembly language, the maximum frame rate is 3 processed frames/sec. Trade-offs of using TV systems with analog processing vs diode arrays with digital processing are discussed.
We discuss the development of a unique optical system intended for capturing drawings to be used in CAD systems. The scanhead consists of two linear rows of optical fiber; one row provides illumination to the drawing surface being scanned, while the other adjacent row transmits the scattered light to a CCD camera for sensing. The analog video signal from the camera is thresholded to binary black/white, noise filtered, and compacted into run-length-coded format for storage as a raster file. The drawing is sequentially stepped and scanned to capture the entire drawing. Our work shows that this system has some unique optical features. In particular, it has characteristics of a transmissive system rather than reflective as might first be expected.
The successful use of deep-depletion silicon CCDs as single photon x-ray detectors is described. CCDs used as x-ray detectors offer the unique combination of high spatial resolution and good non-dispersive spectral resolution. Deep-depletion CCDs additionally offer significant improvements in x-ray quantum efficiency. At an x-ray energy of 5.9 keV, we have measured a quantum efficiency of 76% for a 56 PM deep, 4000 ohm-cm RCA CCD, more than three times that of a conventional CCD (10 pm, 10 ohm-cm). Our tests with this RCA deep-depletion CCD have demonstrated good charge transfer efficiency (0.9999), at an operating temperature of -80 C. Dark current generation is -0.4 e/pixel/s at this temperature, sufficiently low for x-ray astronomy applications. Results of optical tests to measure system gain and CCD readout noise are also presented and discussed. Because the amount of charge per event generated by minimum ionizing cosmic rays in a deep-depletion CCD is > 5 times that of a conventional CCD, energy discrimination can be used to reject -90% of the charged particle background in a deep-depletion CCD. The high x-ray quantum efficiency, high spatial resolution, and ability to reject charged particle background make the deep-depletion CCD one of the strongest contenders for focal plane instrumentation on AXAF, the Advanced X-ray Astrophysics Facility.
The status of laboratory and telescope tests of integrated infrared detector array technology for astronomical applications is described. The devices tested represent a number of extrinsic and intrinsic detector materials, and various multiplexer designs. IR arrays have now been used in successful astronomical applications. These have proven that device sensitivities can be comparable to those of discrete detector systems, and that excellent astronomical imagery can be produced.
For the past several years, Jet Propulsion Laboratory (JPL) personnel have worked with Cincinnati Electronics to develop high-performance infrared line arrays using photo-voltaic Indium Antimonide coupled to a Peticon MOS-switch multiplexer. The result is a high-performance integrating detector which has been demonstrated with integration times up to one hour at 46K.
For several years, the R&D group at KPNO has been evaluating the performance of a 32 X 1 InSb CID array obtained from General Electric (Syracuse) under background conditions anticipated for ground-based astronomical spectroscopic applications. The array, consisting of 100 pm square pixels on 125 pm centers, is intended for use in a cryogenic grating spec-trometer over the range 1.3 - 5.2 pm, with resolving power X/n -100 - 1000. To summarize the results of this evaluation12 i , it was found that an operating temperature -15 K would permit the MOS silicon scanner to operate reliably (but with changes in procedure from 77 K operation) and reduce the thermal dark current to negligible values. Quantum efficiencies measured in the laboratory under these operating conditions ranged from -0.8 at 2.2 pm to 0.25 at 4.7 pm. The most serious problem encountered with this array is the significant delay in output response to incident flux under conditions of low background. This "response lag" could present linearity problems in spectroscopic applica-tions where the integration time is limited by a strong spectral feature or by the thermal background at the long wavelength end of the spectral region being measured. In the summer of 1983, the array was sent to GE for repair of the scanner; in the course of this repair, four dead pixels were removed from the readout sequence and additional CID pixels at one end of the array were connected to the scanner. Thus, the new 33-element scanner reads out a linear array subtending 37 pixels, with pixels 4, 15, 19, and 21 missing from the sequence. After this repair was completed, the array was mounted in a cryostat for an evaluation of its performance as an imager through broadband photometric filters. This paper will summarize the imaging test and further experiments on the response lag problem and will conclude with some initial results with the array operating in the cryogenic grating spectrometer.
A reject device from the NASA/JPL development program for the Space Telescope has been evaluated for its utility in imaging at extremely low flux levels. This thinned, backside-illuminated 800 x 800 imager is run at cryogenic temperatures, allowing long integration times on faint astronomical sources. The device has to date been successfully used for both high-dispersion echelle spectroscopy and for low-dispersion work with our Faint Object Grism Spectrograph (FOGS). In both these applications the T.I. 3-phase imager has shown exceptionally good performance compared to other CCD's.
The STARLAB satellite telescope will use four advanced high performance detectors, operating on the Intensified Photon Counting CCD Array principle. The operation of the detectors is summarized. The baseline design is presented. This design has to meet stringent electro-optic performance specifications while addressing the electronic & mechanical requirements of a large unit operating in space. The tasks being undertaken to develop the detector are outlined, including design, construction & test of prototype detectors, the development of a computer simulation of the detector & supporting technology studies. The detector has the flexibility to be applied to a variety of ground-based & space projects.
The CCD is a thinned, rear-illuminated, RCA device with 320 rows and 512 columns of which about 316 x498 are photometrically useable. The array is over-clocked by the equivalent of 20 extra rows to provide a d.c. bias reference. The array has exceptionally high spectral band width and responsive quantum efficiency and operates in an evacuated cold housing in an analog integration mode at -125°C. On-chip amplifier 'glow' has been limited by reducing drain voltages during integration. The average system read-out noise/sensing element (pixel) is 71 electrons rms. The system response is linear from about 300 electrons to the full-well potential of 6 x105 electrons. There is a significant roll-off in charge transfer efficiency at low signal levels which is probably due to non-overlapping of the gate-electrode structure. (Not a problem with more recent versions of this device). The transfer efficiency was optimized by increasing the overlap and the amplitude of the clock transitions. About 30 pixels have low sensitivity and one third act as 'traps' by attenuating subsequent charge transfers from their column. A few pixels 'glow' causing signal saturation around them. The CCD can be calibrated with a pixel to pixel reliability over the whole array of 2% using illumination of the dome interior as a flat field and to <1% using dark fields in the night sky. Optical fringing from night sky lines is not important in B and V, and can probably be calibrated at longer wavelengths.