Charge-coupled devices (CCDs) are presently the technology of choice for most imaging applications. In the 23 years since their invention in 1970, they have evolved to a sophisticated level of performance. However, as with all technologies, we can be certain that they will be supplanted someday. In this paper, the Active Pixel Sensor (APS) technology is explored as a possible successor to the CCD. An active pixel is defined as a detector array technology that has at least one active transistor within the pixel unit cell. The APS eliminates the need for nearly perfect charge transfer--the Achilles' heel of CCDs. This perfect charge transfer makes CCD's radiation 'soft,' difficult to use under low light conditions, difficult to manufacture in large array sizes, difficult to integrate with on-chip electronics, difficult to use at low temperatures, difficult to use at high frame rates, and difficult to manufacture in non-silicon materials that extend wavelength response. With the active pixel, the signal is driven from the pixel over metallic wires rather than being physically transported in the semiconductor. This paper makes a case for the development of APS technology. The state of the art is reviewed and the application of APS technology to future space-based scientific sensor systems is addressed.
The application of a neural network for processing of the output signal of an infrared (IR) emission mode spectrometer is investigated in this paper. A set of spectral patterns representative of 16 different compounds has been simulated using normal distribution line profiles. These data have then been combined with atmospheric transmittance and path radiance. The network, following training, has been presented with a test set consisting of perturbed versions of the spectra. Perturbations analyzed were line width, peak height, and center variations. The last effect is due to slit image curvature caused by a finite length slit in our hypothetical spectrometer. The purpose of the atmospheric and optical analysis was to insure a realistic estimate of phenomena expected in a field application. The network was found to recognize the input patterns correctly over a broad range of perturbation parameters. We propose that once a satisfactory set of connection weights is established, these should be transferred to a parallel processor (electronic or optical). The network considered in this paper proved capable of generalization under all but the most extreme conditions. Such performance allows by passing of intermediate signal processing for spectral analysis. Consequently, this sort of a system would form a fast and accurate spectral recognition instrument capable of operation under unpredictable conditions.
Photosensitive elements with well-chosen geometry, combined with suitable analog and digital circuitry on the same CMOS/CCD chip, lead to 'smart image sensors' with interesting capabilities and properties. All our smart sensors were fabricated with commercially available multi-process wafer services of CMOS process, one of them with a buried-channel CCD option. Measurement of the optoelectronic properties of standard CMOS/CCD processes (wavelength-dependent quantum efficiency, lateral homogeneity of quantum efficiency/photo- conductivity, CCD charge transport efficiency, etc.) show excellent performance. The smartness that lies in the geometry is illustrated with a single-chip motion detector, a 3-D depth video camera, a single-chip planar distance sensor, and a sine/cosine (Fourier) transform sensor for fast optical phase measurements. The concept of problem-adapted geometry is also shown with a dynamic frame-transfer CCD whose pixel size and shape can be changed electrically in real-time through charge-binning. Based on the wavelength-dependent absorption of silicon, all-solid-state color pixels are demonstrated by properly arranging the available pn-junctions in the third (bulk) dimension. Moderate color measurement performance is achieved using an unmodified CMOS/CCD process, with a CIE general color-rendering index of Ra equals 69.5.
The design and projected performance of a low-light-level active-pixel-sensor (APS) chip with semi-parallel analog-to-digital (A/D) conversion is presented. The individual elements have been fabricated and tested using MOSIS* 2 micrometers CMOS technology, although the integrated system has not yet been fabricated. The imager consists of a 128 X 128 array of active pixels at a 50 micrometers pitch. Each column of pixels shares a 10-bit A/D converter based on first-order oversampled sigma-delta ((Sigma) -(Delta) ) modulation. The 10-bit outputs of each converter are multiplexed and read out through a single set of outputs. A semi- parallel architecture is chosen to achieve 30 frames/second operation even at low light levels. The sensor is designed for less than 12 e- rms noise performance.
A 640 X 400 pixels and 64 dots/mm2 2-dimensional (2-D) contact image sensor has been fabricated by integrating amorphous silicon (a-Si:H) photodiodes (PDs) and thin film transistors (TFTs). The sensor consists of 400 gate and 640 signal lines, and thus contains 256,000 pixels. The PD has a Cr/n-i-p a-Si:H/ITO structure. The TFT has an inverted staggered structure and Ta/TaMo is used for a gate electrode in order to suppress a gate pulse delay. When the TFTs of the n-th row are turned on, photo-generated charges stored in the PD capacitances in the n-th row are transferred to their respective signal line capacitances. After that, the charges are detected by an external voltage sensitive amplifier. The 640 parallel outputs are converted to the serial analog signal by a multiplexer for image processing. The sensor has achieved photoresponsivity of 7.2 V/lx-s, photoresponsivity non-uniformity of +/- 8% and the signal-to-noise (S/N) ratio of 50 dB at the operation of 30 ms/frame scanning speed. Crosstalk as the influence of adjacent lines both X and Y direction is less than 1.3%. This value is estimated to enable achievement of more than 64 levels of gray. The reproduced image quality regarding resolution was good for 8-point kanji characters. This technology will be applicable for multifunctional input/output device mounted on a system such as a pen computer and X-ray detector coupled to a scintillator.
This paper describes the performance of a high-speed PtSi infrared charge-coupled device (IRCCD) camera system for recording periodic and transient events in the infrared. Integration times down to 10 microsecond(s) and the ability to record frames based on an external trigger allows this setup to achieve high effective frame rates. The use of a personal computer (PC) as a controller for CCD clocking and video capture results in a flexible camera system design. Examples of the camera system's capacities for short integration time imaging and triggered image capture are presented.
We have modeled the performance of x-ray crystallography detectors by combining the characteristics of the detector and the experimental conditions. From this model, we have derived a single expression, the experimental detective collection efficiency (XDCE), which predicts the detector's performance. This expression is evaluated for detectors constructed from a square array of identical modules, each module consisting of a fiberoptic taper with a phosphor x-ray converter deposited on the large end and a CCD bonded to the small end. Using this expression, we have developed a design for a modular detector. In order to explore parameters of this design, we have constructed a test detector module in which we can change the fiberoptic taper, phosphor converter and CCD. We have measured the DQE, spatial resolution, response linearity, and dynamic range for the test module for a 3:1 taper. From these measurements, we predict the performance of this type of detector for x-ray crystallography.
A thinned CCD QE model derived by Morley Blouke provided more than an estimation of quantum efficiency. The derivation of the model also produced charge flux equations in the direction normal to the imaging surface. First, these equations were converted into velocity expressions and subsequently into velocity profiles along the normal axis. Such profiles are excellent tools for visualizing the charge collection dynamics as a function of device structure. Profile by-products include dark layer thickness and speed of response (collection). The latter characteristic essentially constrains transverse charge spread. A simple random walk model was created in order to evaluate the extent of this spread in terms of radial rms distances. This distance is a measure of point resolution and is expressed as a function of CCD design and incident wavelength. The theoretical radius of charge spread from a point source ranges from about 3 micrometers for near-IR to about 20 micrometers for near-UV light. This article also discusses justifications for the model, although the final justification must wait for funding of experimental work.
At present we are in the transition era from Standard-Definition Television towards High- Definition Television (HDTV). HDTV offers high-resolution images, sound of compact-disc quality and a wide-screen image format. In Europe, two interim transmission systems (D2- MAC, PALplus) are in use/under development for Extended-Definition Television (EDTV), already offering a wide-screen 16:9 image format with improved image quality. We present the first EDTV image sensor which can switch electronically between a 4:3 and a 16:9 aspect ratio. The introduction of an aspect ratio of 16:9 requires specific pixel dimensions in CCD image sensors. Optimal dimensions mean a trade off between light sensitivity and resolution. This trade off is discussed in detail for the EDTV imager. We have been able to achieve EDTV performance on a 1/2' lens compatible image sensor.
The thermally induced charge (dark current) mechanism in CCD's gives rise to a Poisson distribution of random charge values in each pixel over the device. In the case of low radiant flux and/or quantum efficiency coupled with long integration times this may produce a large number of pixels with values significantly above or below the expected 'average' values. Such pixels in isolation usually pose no significant problem, but may be subject to misinterpretation if randomly aggregated (clustered). In many cases this is of little concern since large quantities of data are captured for subsequent analysis and these random occurrences will be recognized as such. But in the case where cost or complexity mandate 'one-shot' data capture, the question of how often such occurrences may be expected is altogether reasonable. A probabilistic model of such clustering is developed and several scenarios evaluated.
A fast centroid processor developed for telescope pointing and high speed tracking is described. It is based on a modified Photometrics CCD camera using a 24 X 24 pixel frame transfer CCD. The small CCD allows very fast update rates (200 frames per second with an 8 X 8 window). It provides good read noise performance (11 e- at 40 kHz read rate) and 16 bit data resolution. A self-adjusting algorithm accomplishes background subtraction. The centroid processor, based on a digital signal processor (Motorola 56001), calculates centroids in real time on 16 bit data to sub-pixel accuracy. The centroid values are converted to 16 bit data words which are used to drive the servos of a chopping secondary mirror on the 2.1 m telescope at Kitt Peak National Observatory. The mirror can track apparent star motion to provide low level (tip/tilt) adaptive correction of astronomical images. A factor of 5 improvement in Strehl ratio is theoretically possible at the 2.1 m telescope with tip/tilt correction. Preliminary results have yielded a factor of 2 Strehl ratio improvement.
The performance of a large format (512 X 512, 20.3 mm diagonal) Charge Injection Device (CID) imager which was fabricated for use in spectroscopy, microscopy and other scientific instrumentation applications is reported herein. The device incorporates a large (28 X 28 micron) pixel size and on-chip signal amplification to achieve large full well capacity, low noise and wide dynamic range. As do other CIDs', the device also features a broad spectral response, virtually no blooming, true Non-Destructive Signal Read-Out (NDRO), video skimming and individual pixel address capability. Together, these unique CID features provide the capability to extend the imager dynamic range, achieve real-time signal monitoring and read-out for adaptive exposure control, and achieve lower noise through NDRO signal averaging.
An image sensor with lateral antiblooming for overillumination protection is discussed. The device is implemented using a double poly, NMOS buried channel CCD process with a buried drain that runs adjacent to the channel stops. The antiblooming barrier is formed by a surface channel region adjacent to the buried data. Typically these devices are implemented using a three poly process but by eliminating the exposure control requirement, a two poly process technology can be used. The area array pixels are built using four phase CCD technology maximizing charge handling capacity. The surface channel antiblooming barrier confines the charge to buried channel operation.
Radiation hardness is critical for charge coupled devices used in the electron bombarded mode. Two types of damage in CCDs are caused by keV electron irradiation: a flatband voltage shift and an increase in interface state density. A flatband voltage shift is more catastrophic to device performance than an increase in interface state density, especially for MPP devices. The type of radiation damage a CCD is susceptible to depends on the process used to fabricate it. Results are presented which show that Tektronix CCDs fabricated with a straight silicon dioxide gate insulator exhibit an increase in interface state density but little if any flatband voltage shift.
A page-sized flat-plate image sensor array intended for use as a contact imager for documents is described. The imaging device is a matrix-addressed pixel array of light sensors fabricated from hydrogenated amorphous silicon on a glass substrate, with each pixel comprising an n-i-p sensor and a thin film switching transistor. The array requires no imaging optics and therefore collects light very efficiently, making it well suited to high speed document input or for medical X-ray imaging. The array fabrication, drive electronics, imaging properties, and applications are discussed, and some examples of images are presented.
Large format charge coupled device area arrays (1 million pixels or more) have proven to be useful in scientific, medical and industrial imaging applications. DALSA has developed a 1024 X 1024 pixel single output, full-frame area array incorporating 3-poly 3-phase buried channel NMOS CCD shift registers and a 10 micrometers X 10 micrometers pixel pitch. The device was fabricated with an additional buried channel implant (notch) in the pixel columns to increase charge storage capacity. In this paper the authors discuss the design and initial performance evaluation of the device. Preliminary measurements of the pixel charge storage capacity indicate 70,000 e- without notch and 140,000 e- with notch. The results indicate that the sensor should be suitable for a variety of applications such as high resolution machine vision, still photography, and scientific imaging.
New low-noise CID imagers are being created to meet the demanding requirements of scientific instrumentation, high-speed tracking and nuclear inspection applications. The imagers incorporate new process technology and/or new low-noise architectures to exploit inherent unique CID features including random pixel addressability, true non-destructive pixel readout (NDRO), two-dimensional windowing (sub-array readout), and exceptional resistance to the effects of ionizing radiation. These CID features enable the user to monitor and dynamically adapt application exposure levels in real-time, reduce noise, and read out small sub-arrays of pixels at exceptionally fast rates. Due to their radiation tolerance characteristics, the devices can operate in harsh radiation environments and actually image (detect) the incoming radiation. Device formats and performance features are summarized.
A 26.2 million pixel CCD Imager Sensor has been successfully designed and fabricated. The device uses a full frame architecture with 5,120 X 5,120 pixels organization. With a pitch of 12 microns in both dimensions, the overall image zone is 61.44 mm X 61.44 mm. The charge storage capacity of each photosite is greater than 130,000 electrons and the minimum detectable charge is 50 electrons when correlated double sampling is used. The device is also capable of reduced dark current operation of 60 pA/cm2 when operated in the surface inversion mode. The device has four outputs, each of which can operate up to 12 MHz.
We have developed an astronomical imaging system that incorporates a total of eight 2048 X 2048 pixel CCDs into two focal planes, to allow simultaneous imaging in two colors. Each focal plane comprises four 'edge-buttable' detector arrays, on custom Kovar mounts. The clocking and bias voltage levels for each CCD are independently adjustable, but all the CCDs are operated synchronously. The sixteen analog outputs (two per chip) are measured at 16 bits with commercially available correlated double sampling A/D converters. The resulting 74 MBytes of data per frame are transferred over fiber optic links into dual-ported VME memory. The total readout time is just over one minute. We obtain read noise ranging from 6.5 e- to 10 e- for the various channels when digitizing at 34 Kpixels/sec, with full well depths (MPP mode) of approximately 100,000 e- per 15 micrometers X 15 micrometers pixel. This instrument is currently being used in a search of gravitational microlensing from compact objects in our Galactic halo, using the newly refurbished 1.3 m telescope at the Mt. Stromlo Observatory, Australia.
A high speed imager capable of 1 gigacycle frame rate was developed with 80 micrometers X 80 micrometers pixels. A special multi-phase readout shift register and parallel to serial transfer structure are designed for high charge transfer efficiency operation across a large 80 micrometers pitch at high speed. A total of 64 outputs each designed to operate at 16 MHz are distributed around the array. In order to minimize power dissipation, a single output MOSFET operating in conjunction with an off-chip transimpedance amplifier was used. A photodiode, interline transfer pixel architecture was used with lateral antiblooming and exposure control capability.
A transparent, semi-solid, electrolytic gate has been applied to the backside of thinned CCDs for quantum efficiency enhancement. The gate is applied by spreading a water solution of phosphoric acid and polyvinyl alcohol onto the silicon and drying it to form a thin plastic film. When a negative voltage of less than one volt with respect to substrate ground is applied to the gate, a QE pinned condition (100% internal quantum efficiency) is produced. An insulating layer is not needed with this gate (as it is with electronic conductors) since a threshold voltage of about 1.2 V is required before conduction into the silicon can occur. The mechanism of charging is believed to involve a pile-up of negative ions at the silicon-electrolyte interface which compensates for the positive oxide charge. Conduction into the silicon at low voltages is restricted by the oxidation potential of the negative ions in the electrolyte.
The optimization of back illuminated CCDs for low-light-level applications requires many process steps. One such step is the deposition of thin films on the freshly thinned backside surface. These films may consist of many layers depending on both the desired properties of the detector and on the backside charging mechanism. We describe our backside coating process which has been optimized for astronomical applications. After thinning, we first grow a thin silicon oxide film in a steam environment. Following oxidation we deposit an antireflection coating optimized for a particular wavelength. We may also deposit a thin film of platinum between these layers that acts to charge the backside. Using these thin film coatings we have been able to produce CCDs which reach silicon's theoretical maximum quantum efficiency over the 300 - 1000 nm wavelength region.
In this paper we report on recent work on the development of a stable ultraviolet sensitive antireflection coating for use on charge-coupled devices. The coatings that are discussed in this paper are single layer Si3N4 and a dual layer Si3N4/MgF2 coating. Both coatings provide excellent quantum efficiencies at 300 nm (> 50 - 60%) and useful response down to 200 nm (the limit of the measurement capability). Ultraviolet (UV) flooding is shown to be effective in increasing the quantum efficiency. The films demonstrate a long time constant decay of the UV enhanced quantum efficiency which is of the order of months at room temperature. A simple model of a thermally activated process to characterize the decay in quantum efficiency has been developed.
X-ray microscopy requires image detectors for soft x-rays (2.4 nm to 4.5 nm wavelength) with high detective quantum efficiency for a low radiation dosage applied to the sample. A thinned backside illuminated CCD has been attached to the Gottingen x-ray microscope, which is installed at the BESSY electron storage ring in Berlin. The CCD was a commercially available device with 1024 by 1024 pixels (each 24 micrometers square) without the anti-reflecting coating, which is applied to the standard device. First experiments performed at the primarily used x- ray wavelength of 2.4 nm show a considerable reduction of exposure time compared to the previously used photographic emulsion. This greatly reduces the radiation dose applied to the sample specimen. There was no degradation in performance of the CCD detected after one week of operation.
A full frame, four million pixel CCD area array has been developed for high performance applications. The device consists of 2048(H) by 2048(V) active elements with four outputs, and is fabricated using a triple poly, buried n-channel CCD process. Design features which enhance array performance include a Multi-Pinned Phase (MPP) pixel design for reduced dark current, low noise output amplifiers, and a phosphor coating to extent the spectral response. Excellent performance is demonstrated with a dynamic range of 93 dB, a read noise of 2.5 electrons (r.m.s.) and dark current of 20 pA cm-2 (25 degree(s)C).
Protons from solar flares represent the major threat to the scientific performance of a CCD in the SOHO orbit at L1, decreasing CTE and thus non-uniformly degrading the MTF of the detector. Lattice damage assessment and prediction rely on accurate radiation damage experiments to 'calibrate' numerical simulations and modeling. The energy ranges where TRIM and NIEL represent valid models overlap around a few MeV. Thus, the proton beam from Lockheed PARL's 0.1 to 3 MeV Van de Graaff generator provides a convenient test facility. We present results from an accurate experiment using 2 MeV protons on the MDI detector (LORAL 1024 X 1024 21 micrometers 3P MPP CCD). A premiere feature in the experiment is the achievement of a stable, uniform low fluence and extremely accurate dosimetry at this relatively low energy. Pre- and post-radiation CTE measurements for our specific mode of operation (relatively fast readout rate of 500 kpix/s) is obtained using Fe55 method over a wide temperature range. They reveal somewhat unexpected results. The damage is more severe to parallel CTE than to serial CTE and the former worsens when cooled down to -50 degree(s)C, then improves when cooled further.