Recent work at the Rome Air Development Center (RADC)1 and elsewhere' has indicated that the performance of sensors based on staring focal plane arrays ultimately may be limited by their nonuniformity of response. It has been shown that simple addition and multiplication operations alone are not sufficient to remove all detector-to-detector response variations. Post-correction error remains, arising from the fact that the individual detector sensitivities do not necessarily differ by a simple constant, but rather can differ in a spectrally dependent fashion across the entire waveband of interest. Using the RADC linear model for a general focal plane array, the theoretical MWIR response of a 256x256 PtSi device is calculated, and the effects of uncompensated imaging are demonstrated. Initial lab testing of this device will employ a single-point (additive) compensation method. Mathematical analysis is presented that predicts the contrast signal-to-noise ratio (CSNR) one might expect with such a scheme. It is demonstrated that even modest amounts of spectral nonuniformity can seriously degrade the ability to resolve small temperature differences in a thermal scene. It is anticipated that actual device testing will begin soon, with the hope of being able to quantify the individual magnitudes of the additive, multiplicative, and spectral nonuniformities of the device.
Large staring PtSi Schottky-barrier focal plane array (FPA) sensors are fabricated using near-standard silicon technology and have represented a practical way to achieve outstanding thermal sensitivity with earth-background imagery. Their application potential could be extended significantly if background-limited operation can be maintained with low-level space backgrounds. This paper will discuss noise, dark current, and readout efficiency performance requirements and present low-noise test data obtained on 160x244-element, 5.5um-cutoff PtSi FPA's operated with exposures 10 times lower than those from typical earth backgrounds. In addition, test data will be presented on developmental IrSi detectors which have a 10um cutoff and have a quantum efficiency at 5um an order of magnitude higher than that of PtSi. This improvment in responsivity combined with low noise performance will permit design of high sensitivity sensors in both MWIR and LWIR bands.
Improvements in CCD sensitivity in the UV and near-IR by using backside illumination, external backside accumulation, and wavelength conversion phosphors is discussed. Quantum efficiencies of greater than 50% in much of the UV and visible are reported through the use of backside illumination and either UV flood or flash gate backside accumulation. Problems with the environmental stability of backside accumulated CCDs and attempts to control the stability through the use of a biased flash gate is discussed. Significant QE gains in the near-IR through backside illumination without backside accumulation are reported. The use of a high-efficiency wavelength conversion phosphor to increase the UV sensitivity of frontside illuminated CCDs is presented. Results are presented that demonstrate that the QE of a typical polyphase frontside illuminated CCD can be increased from virtually zero to 20% in the UV.
We present results of a program to enhance the ultraviolet and extreme ultraviolet response of charge-coupled devices (CCDs). The ultimate goal of our program is to develop a large format device with both high and stable quantum efficiency from 100-3000 Å which can be used as a windowless imaging detector in a space environment. Of particular concern in this environment and at these wavelengths is long-term photometric stability under hard vacuum. We will discuss quantum efficiency and vacuum stability measurements for a number of ion implanted and laser annealed test CCDs specially fabricated by Tektronix for this program.
Methods to improve the red and near-IR quantum efficiency (QE) of CCDs are discussed. The effects of thinning, antireflection (AR) coating, and mounting CCDs for use in astronomical observations at wavelengths greater than 700 nm are presented. Thinning CCDs to inside their epitaxy and backside illuminating them increases their QE over thick devices even in the far red, due to the absence of frontside absorption and reflection losses. AR coatings decrease the backside reflection loss and greatly reduce the amplitudes of interference fringes. Mounting CCDs on a substrate eliminates surface warpage which improves the user's ability to correct for instrumental and background signatures. Some of these optimization techniques are demonstrated using the Photometrics PM512 CCD made by Ford Aerospace.
Low frequency noise with power spectra of the form 1/f' has been investigated for an n-buried channel CCD by measuring the drain-to-source noise voltage. When VG = 0 to -1 V, the device is in the surface-channel mode and the power spectra of the noise exhibits 'y = 1. As the gate voltage decreases, the noise for f < 50 Hz increases and eventually peaks at VG = -2.6 V before it decreases as the device moves into the buried-channel operation. On the other hand, the noise for f > 50 Hz peaks twice at VG = -2 and -3 V. Between the two peaks, there occurs a dip at VG = -2.4 V where -y takes a maximum value of 1.7. The preliminary results indicate that the noise is primarily due to tunneling of electrons at the silicon-silicon dioxide interface to traps located inside the oxide. The traps farther from the interface affect the noise for f < 50 Hz whereas those nearer to the interface influence the noise for f > 50 Hz.
Architectures for focal-plane image processing using CCD circuits are discussed. The choice of architecture depends on imager density and required throughput. High-density imagers require the reconstruction of local neighborhoods prior to image processing. Lower density imagers can utilize spatial parallelism to improve throughput. The use of three-dimensional structures can provide additional real-estate for processing circuitry.
Automated vision applications including printed circuit board inspection, parts measurement, and article positioning and tracking rely upon accurate determination of object edges. With most solid state sensors this is limited by the actual sensor resolution (ie. total number of pixels). Sub-pixel interpolation can be performed on the video output, but the accuracy of such calculations is highly dependant upon the imager pixel topology. The large active (light sensitive) area and contiguous nature inherent to the CID pixel structure is well suited to sub-pixel interpolation techniques. As a result, higher effective resolution can be obtained from an existing format CID sensor. The CID pixel is examined. Factors affecting the accuracy of this interpolation are identified. Estimates of the sub-pixel interpolation capability of some CID cell configurations are provided.
Two varieties of Focal Plane Convolvers have been fabricated; the time multiplexing and the space multiplexing focal plane convolvers. They extract spacial features of a scene using parallel processing procedures at the imaging detector array site. The detailed description and operation of these devices is described. The following related areas of technology are also discussed: o Double Correlated Double Sampling o CCD Charge Transport o Equal Variance Mapping o Charge Transport Efficiency o Spacial Convolution Operators o Uniformity Issues o Local Rotational Invariance o Other Devices on the Reticle
The application of GaAs acoustic charge transport (ACT) technology to the development of high speed charge read-out multiplexers for imaging systems is described. Barrier storage cell structures are developed to permit extended time integration of signal charge directly in the ACT channel while inhibiting SAW charge transport. These storage cells, coupled with lateral ACT channel charge injectors and a GaAs MESFET based non-destructive charge detec-tion circuit form the basis for a focal plane compatible multiplexer technology with high read-out rate and large dynamic range. The high performance charge transfer characteristics and focal plane compatibility obtained from cryogenic ACT device operation are demonstrated. The basic structure, func-tional principles and experimental operation of a single channel 64 cell developmental ACT multiplexer operating at 360 MHz SAW clock rate are presented. The work indicates that ACT technology has significant potential for high speed image read-out applications.
Progress in high-speed GaAs charge-coupled device (CCD) research is described. Experimental and modelling results are reported for two different structures; capacitive gate CCD's and resistive gate CCD's. A charge packet replicator/subtractor circuit is discussed.
The design and performance of a 1024x1024 pixel charge-coupled device (CCD) imager are described. This device is fabricated utilizing a 3-phase, three-level polysilicon gate process. The chip is thinned and is employed in the back-illumination mode. Detailed measurements including imagery, read noise, full well capacity, charge transfer efficiency, linearity, dark current, spectral response, residual image, and charge collection efficiency are reported.
A new device architecture was developed for building high-performance and high-resolution image sensors suitable for consumer TV camera applications. The sensor elements employed in this architecture are junction field-effect transistors that are organized into an array with their gates floating and capacitively coupled to common horizontal address lines. The photogenerated signal is sampled one line at a time, processed to remove the element-to-element nonuniformities, and stored in a buffer for subsequent readout. The described concept, which includes an intrinsic exposure control, is demonstrated on a test image sensor that has an 8-mm sensing area diagonal and 580(H) x 488(V) picture sensing elements. The key performance parameters achieved in this design, in addition to a high packing density of sensing elements with a unique hexagonal shape, include high signal uniformity, low dark current, good light sensitivity, high blooming overload protection, and no image smear.
A physical model is used to describe the image modulation and noise characteristics of charge-coupled device (CCD) imagers. The model includes the effects of photon absorption, charge generation and collection, and signal quantization. The imaging characteristics are presented in terms of the MTF and Wiener, or noise power, spectrum. A computed example illustrates the use of the model.
For most thinned silicon CCDs, the photosensitive volume is bounded on top and bottom by layers of silicon dioxide. The frontside oxide is grown to serve as an insulator beneath the conductive gates of the parallel array while the backside oxide forms naturally as the initially bare silicon oxidizes. This paper describes the characteristics of the interface between these oxides and the photo-sensitive silicon and indicates the extent to which CCD performance (e.g. dark current, spectral response, charge collection efficiency, charge transfer efficiency, pixel-nonuniformity read noise full well capacity blooming residual image and vulnerability to ionizing radiation damage) depend* upon these interfacial characteristics. Techniques are described to achieve optimum passivation of these interfaces and to thereby obtain superior performance in the areas just listed. Specifically an implanted structure (the Multi-Pinned-Phase, MPP) is described which provides excellent frontside passivation and several techniques (backside charging, flash gate, the biased flash gate and ion-implantation) are presented for back surface passivation.
This paper describes the design and performance of a scientific CCD array for use in astronomy applications. The device is a four-phase, buried-channel CCD structure that operates in the frame-transfer mode for imaging the scenes at low light levels in the spectral regions from 450 nm to near infrared. The sensor consists of two sections of 1200 x 200 CCD imaging elements with 27 x 27 μm pixel size. Each section can be operated independently with its own four-phase vertical clock or as a full frame 1200 x 400 imager when both sections share the same four-phase clock. The device' serial-parallel-serial structure allows the charge in the parallel (vertical) CCD shift registers to be able to shift either up or down to the top or bottom serial (horizontal) CCD readout register. The readout register, in turn, transfers the charge in sequence to the output amplifier. The high full-well capacity of each pixel (> 0.6 million electrons) and low readout noise (< 4 electrons rms) yield a dynamic range of more than 103 dB. Using the multi-pinned-phase (MPP) mode of operation, the saturation time is more than 30 seconds at room temperature. The device has an excellent charge-transfer efficiency (CTE) of more than 0.999996 at low light level and does not exhibit any charge-packet problem which has plagued other CCD manufactures. In addition to top-side illumination, the device layout also can accomodate back-side thinning and back-side illumination for improvement on short wavelength spectral response. Subject terms: astronomy; charge-coupled device (CCD); near-infared; full-well capacity; multi-pinned-phase (MPP); dynamic range.
A high-speed MOS imaging array prototype has been designed, fabricated, and tested. Preliminary tests of the 64 x 32 pixel device show imaging operation at more than 1000 frames per second. A 256 x 256 pixel array projected to achieve 2000 frames per second is under development. The sensor will be designed into a RAM based camcorder capable of storing 2000 frames of data. The camcorder will measure approximately 6" x 6" x 12" and contain 128Mbytes of dynamic RAM and 64 flash A/D converters. Camcorder outputs include standard video and a digital port.
The four quadrant CCD was developed to meet the reliability needs of future space programs that use large area CCDs. The 10242 format was selected as a precursor to the 20482 format. An experimental lot of 10242 CCDs was made in 1987 and a production lot was run in 1988 for the AXAF star tracker program. Excellent performance was obtained with the production lot and the four quadrant operation was demonstrated to work as predicted.
An all-parallel output CCD imager was developed for high frame-rate imaging applications. This imager was designed for 5000 frames per second but has been operated at much higher frame rates. The all-parallel output design was chosen to keep the pixel rate for each output low and thus obtain low-noise performance. A frame-transfer architecture was implemented to obviate the need for a fast shutter system. The device can be implemented using a backside-thinned CCD technology to achieve an improved blue response.
Detailed measurements and characterization of trap behavior in large area CCDs have been performed. The effect of these defects is to cause a local decrease in the charge transfer efficiency in the affected pixel. Bias-temperature stressing of the device has lead to the conclusion that the source of the traps is in the dielectric. A model is proposed which can explain all the current data.
A partially-inverted mode of operation of the Tektronix 10242, 3-phase CCD has been tested and characterized. In this mode, two of the three parallel phases are held in the inverted mode during integration while the third is held in the non-inverted mode. This mode offers a compromise between the low dark current and low full well capacity of fully inverted operation, and the high full well and high dark current of non-inverted operation.
The Kodak SV9600 Still Video Transceiver is designed to electronically transmit and receive high quality video images over standard telephone lines. The transceiver captures a full frame of video, digitizes and stores it in memory for manipulation and display of the image data. The data is compressed using a highly sophisticated algorithm developed at Kodak. The compressed image data is transferred to a built-in modem for high-speed communication over standard telephone lines.
The Kodak still video transceiver system is designed to electronically transmit and receive high quality color video images over standard telephone lines. A detailed description of the algorithm used in compressing the digital image data is provided. The algorithm is based on the discrete cosine transform (DCT) and uses human visual system (HVS) sensitivity models to achieve high compression ratios without visible artifacts.
A color scanner producing high-resolution digital image data, 4096 x 6144 pixels per color from 35 mm color negative film, has been developed. A linear charge-coupled device (CCD) image sensor and mechanical translation are utilized to achieve 4425 lines per inch sampling pitch. Red, green and blue fields are scanned sequentially, each in 34 seconds. At 1.0 megapixel/second, the measured data are digitally processed to remove individual photosite variations and are transformed to log space. The processing also includes data averaging; this option provides lower spatial resolution data at higher signal-to-noise ratios. Voltage signal-to-noise ratios of 2400:1, or 67 dB, have been achieved at the highest resolution. This paper discusses the design and performance of the scanner. Measured versus computed spatial frequency response is presented. The hardware processing is described. A model for processing computational error analysis is presented along with noise measurements. Finally, special film considerations are discussed.
Line scan CCD imagers have been used successfully in various applications such as remote sensing, document and film scanning, industrial inspection, and optical processing. However, high quality scanning of transparencies for electronic color separation has not been completely satisfactory when performed with CCD imagers intended for document scanning applications. This paper describes a new CCD imager specifically designed to handle the wide dynamic range of transparencies. This new line scan CCD imager features 6000 pixels 10μm x 10μm on a 10μm pitch. The 2.4 inch long chip is built on a low crosstalk substrate with all the state-of-the-art features: low-lag photodiodes, antiblooming and integration controls, output clock gates, in-phase outputs, and high scale factor amplifiers. The device is packaged in a new ceramic/precision metal header. A key performance parameter is its dynamic range: 30,000:1 (relative to RMS temporal noise in the dark, with CDS) which makes it an excellent candidate for use in high-resolution high-quality flatbed scanners.
COD linear image sensors can fulfill several functions in professional TV equipment. For HDTV, the required characteristics imply the development of new products. Whithin the European Eureka program, THOMSON-CSF has designed a 2048 element image sensor to be used in telecine equipement. Improved vertical transfer efficiency together with a data rate of 60 mega pixels per second are the two main features of the device.
A high precision exposure control system using a frame interline transfer (FIT) CCD image sensor and a compact aperture mechanism is proposed for consumer still video cameras. In this system, an electronic shutter, which is a feature of the FIT CCD image sensor, is used instead of a mechanical shutter. This electronic shutter can be used with flash. The aperture mechanism is constructed of three different size apertures and two solenoids for aperture selection. The experimental results show that the exposure error of the entire system is 0.15 EV or less.
It is more than seven years since the advent of the electronic still video (SV) cameraY)It gave a pretty big impact to the conventional film and camera-relating industry at first The market of still video cameras, however, did not grow as rapidly as expected. So far, only a commercial market has been created, where instantness and ability to display on TV screen and to transmit by telephone cable are considered to be important. For a new consumer market, not only above mentioned advantages, but also better picture quality, smaller camera size, and lower price are reqired. Those requirements have been fullfilled step by step by improving CCD and signal processing circuit, and production technology. Here, we tried further improvement of picture quality of still video camera by developing a digital still video camera.
A discussion of media recording specifications as they are defined by the still video floppy standard will be used to lead into the type of head and me interface used on still video floppy products in the marketplace. The Kodak interface consist of the following components: dual thin film head, floppy disk media, and the button interface (media stabilizing surface). By following the development of Kodak's button/head interface, an insight into the problems encountered in the interface's progress from the development stage to the final product design will be obtained. Along the road to the production design, many changes had to be made in order to insure the integrity of the interface. Discussion will elaborate on the problems encountered with materials, surface coatings, currents am, static, and manufacturing techniques. A major yardstick by which the interface performance was evaluated consisted of the capture between the media and the head/button interface. This parameter was evaluated by what we define as e capture window, a plot of data points relating the tilt, attack, penetration, and signal level of the button/head interface to the media. The life of the interface's components is a critical performance parameter. A discussion of the typical life of the interface components will also be presented.
A method used to develop an optimized design for a single-sensor color electronic camera is described. A color filter array (CFA) that senses red (R), green (G), and blue (B) information in a manner consistent with the relative sensitivities of the human visual system (HVS), an optical prefilter, and an optimized image reconstruction scheme are utilized in this system which maximizes image quality for a given number of pixels on a sensor.