This paper presents the rationale, based on sensitivity and reduced optical aperture, for selecting a staring sensor over a scanning sensor for space applications where a weak target signal must be extracted from a heavily cluttered background. Performance requirements are also discussed, including parameters of material, sensitivity, uniformity, dead space, resolution, and dynamic range. The important processes in signal processing are presented: filtering, correlation, detection, and tracking.

The Teal Ruby Experiment (TRE) sensor presents unique problems to radiometric performance testing and calibration of a mosaic infrared sensor because of the large number of resolution elements; the wide range of spectral, temporal, and flux level operating regions; and the cryogenic operating conditions. This paper contains a summary of the Teal Ruby test facilities and requirements at the infrared charge-coupled device (IRCCD) detector array, zone assembly, focal plane assembly, and sensor levels. Automated test facilities and capabilities are presented to highlight the development requirements and approaches to testing. Key issues concern the complexity of testing, selection of test parameters, commonality of test algorithms and data presentation, data needs for acceptance testing, optimization and integration, and test equipment standards for accuracy, operating range, and contamination control.

The versatility of Pushbroom Sensor Designs is outlined. Multiple line arrays of detectors are used in the focal plane to provide signal enhancement and track association data format for signal conditioning. Strong targets may be sensed without moving parts. Weak targets are sensed against earth background with a minimum of motion for alignment with a variable image motion direction. Very wide fields of view are covered by multiple telescopes and focal planes to illustrate a conceptual approach for meeting exceptionally wide swath coverage requirements. The options provided by using multiple line arrays of detectors in the pushbroom mode give unique capability to the sensor design.

Deep-space surveillance missions have imposed severe demands on existing technology and simulated the search for new, advanced technology developments to provide higher performance. Defense Advanced Research Projects Agency (DARPA) sponsored Teal Amber as a visible charge-coupled device (CCD) and associated focal plane signal processing technology development and demonstration program. This paper describes this large-scale, staring-array-sensor concept. The current state of art in the resulting visibled CCD imagers is specified, along with the focal plane signal processor implementation in low power-weight-volume large-scale integrated (LSI) circuitry. Performance requirements and analytic predictions are compared to demonstration system results from an electro-optical test site in White Sands, New Mexico.

Staring infrared imagers typically exhibit large d.c. offset level variations and responsivity variations from pixel to pixel. In order to extract the scene information from the focal plane output signal, this characteristic fixed pattern noise must be normalized prior to display. Conventional techniques for this compensation involve the use of a uniform thermal reference source which is periodically introduced into the sensor field of view to act as a calibration of the offset and responsivity variations for each pixel. This viewing of a thermal reference source generally involves use of electromechanical or electro-optical shutters which detracts from the mechanical simplicity of the staring imager. A real-time compensation technique has been developed which utilizes the infrared (IR) scene itself for calibration and continually updates the compensation coefficients without the use of a thermal reference source or shutter. This "shutterless" compensation technique makes use of scene dynamics, averaged over a period of time, as an effective uniform reference source. The results of real-time simulations of this technique have been demonstrated using both FLIR and visible imagery. Results of these simulations are presented along with a discussion of applicable areas for this technique and approaches for real-time hardware implementation.

Infrared (IR) staring arrays offer significant performance improvements over mechanically scanned systems, if the signals from these focal planes can be conditioned for use by imaging displays or signal processors. IR staring arrays offer the potential for increased sensitivity, wide dynamic range, and short frame time, if the problems associated with the readout of data from these arrays can be solved in an efficient manner. Some of the func-tions included in this signal conditioning are: nonuniformity compensation, local area gain and brightness control, detector integration time control, and multiple frame composition. Nonuniformity compensation and local area gain and brightness control were covered in earlier papers. 1,2,3 This paper deals with integration time control and the resulting multiple frame composition possible with focal plane arrays.

32 x 32 and 64 x 64 staring infrared focal plane arrays with epitaxial HgCdTe and InAsSb photovoltaic detectors coupled to surface channel CCD multiplexers have been fabri-cated and characterized. The backside-illuminated detector arrays have bandgaps suitable for operation in either the 3-5 or 8-12 μm region. A source-coupled input circuit with background suppression is utilized. Performance characterization of the multiplexer will be given at cryogenic temperatures. CCD charge transfer efficiency and MOSFET threshold voltages have been measured as functions of temperature. The input circuit 1/f noise will be characterized. The detectivity and noise equivalent temperature of the hybrid focal planes will also be discussed.

This paper is the first of a series 1,2,3 describing a range of efforts at Hughes Research Laboratories, which are collectively referred to as "Three-Dimensional Microelectronics." The technology being developed is a combination of a unique circuit fabrication/packaging technology and a novel processing architecture. The packaging technology greatly reduces the parasitic impedances associated with signal-routing in complex VLSI structures, while simultaneously allowing circuit densities orders of magnitude higher than the current state-of-the-art. When combined with the 3-D processor architecture, the resulting machine exhibits a one- to two-order of magnitude simultaneous improvement over current state-of-the-art machines in the three areas of processing speed, power consumption, and physical volume. The 3-D architecture is essentially that commonly referred to as a "cellular array", with the ultimate implementation having as many as 512 x 512 processors working in parallel. The three-dimensional nature of the assembled machine arises from the fact that the chips containing the active circuitry of the processor are stacked on top of each other. In this structure, electrical signals are passed vertically through the chips via thermomigrated aluminum feedthroughs. Signals are passed between adjacent chips by micro-interconnects. This discussion presents a broad view of the total effort, as well as a more detailed treatment of the fabrication and packaging technologies themselves. The results of performance simulations of the completed 3-D processor executing a variety of algorithms are also presented. Of particular pertinence to the interests of the focal-plane array community is the simulation of the UNICORNS nonuniformity correction algorithms as executed by the 3-D architecture.

A microprocessor based compensation electronics system has been developed in conjunction with a video image display for 64 x 64 (Hg,Cd)Te hybrid staring arrays. The system has a data throughput rate of 5 MHz, where offset (12 bits resolution), gain (8 bits), and defective cell corrections take place in real-time. While imaging in real-time, the system also compiles quantitative data on the array performance such as responsivity and noise in a separately allocated memory. A general purpose computer can be interfaced to this systen to extract array data and perform statistical computations. The system can readily be expanded for staring arrays of 256 x 256 elements without major modifications.

The use of solid-state detector arrays which operate in a pushbroom scan mode for remote sensing of the Earth's resources and environment has received increased attention in the last several years. The potential for improved radiometric sensitivity, geometrical accuracy and signal processing permit consideration of new levels of sensor performance in the areas of spatial resolution, spectral resolution, mapping accuracy, and improved system throughput of data products. These benefits depend on the ability to manufacture and accurately align thousands of detectors into a multispectral focal plane. Two key performance goals are to achieve: (1) radiometric calibration to 0.5% precision detector-to-detector over a dynamic range of 1000:1; and (2) geometric alignment to place the detector elements to within 0.1 resolution element of their desired perfect positions. System cost (including the ground segment) and complexity should be traded against these goals. Science experiments continue to be needed to establish the tolerances on these goals.

This paper is primarily a review of two recent focal plane development projects. The first focal plane consists of six 1024x64 element TDI image sensor chips mounted on a prism beam-sharer. This focal plane is designed to give a video signal with high SNR over a wide dynamic range. The second focal plane consists of a single 1024x128 element TDI image sensor chip. It is designed to perform over a wide dynamic range extending down to starlight scene illumination levels. Both chip designs employ buried channel, silicon gate CCD technology. The silicon gate sensing area is designed for maximum broadband quantum efficiency over the range 500-900nm; an average quantum efficiency over that range of 50 percent has been achieved. The sensor element in both chip designs is a 20x20 micrometer square stage of a four-phase CCD register. This element design provides high MTF and a saturation charge level of approximately 1x106 electrons/pixel as is needed for achieving high SNR. On-chip preamplifiers are designed for specific requirements of the focal planes. The chip for the first focal plane utilizes a highly linear resettable floating gate amplifier; the linearity facilitates matching of imagery generated by neighboring chips. The chip for the second focal plane has two preamplifiers which may be read out in parallel; the first handles the full dynamic range of the CCD output and the second, because of higher gain, handles only the low end of the CCD output. The high gain amplifier has operated at 1MHz sample rate with a noise equivalent signal level of approximately 20 electrons/pixel RMS.

Hg_l__xCd_x/cTe epitaxial layers have been successfully grown in various compositions, for 1-3 μm applications. n⁺/p junctions are formed either by a standard B-implantation into as-grown p-type layers or by doubly grown p- and n-layers. The SWIR HgCdTe photodiodes exhibit quantum efficiencies of 55-65% without AR coating. For the diodes with 1.39 μm cut-off at room temperature, the zero bias detector resistance-area (R_οA) product is 4 x 10⁴ Q-cm², and the dark current density is - 1 x 10^-4 Ω-cm² at half-breakdown voltage. The same values of - 104 Ω-cm² R_oA products have also been measured for 2.4 μm cut-off photodiodes at 195K. The energy gap and temperature dependence of RoA product is in excellent agreement with the bulk limited generation-recombination model. The breakdown voltages of SWIR diodes vary from 12 volts to greater than 130 volts, depending on the Cd composition (x) and base carrier concentrations.

This paper describes a technique whereby the video data from solid state arrays can be calibrated by monitoring the statistics of the observed imagery on a continuous basis. Convergence is shown to be relatively rapid and consistent. The major advantages include the elimination of special requirements for calibration hardware and the need to provide separate calibration time cycles.

A direct gate coupled input circuit between a 1.0 to 2.5 micrometer photovoltaic mercury-cadmium telluride detector and a CCD multiplexer is proposed for low background applications. Theory predicts a D*λpeak on the order of 10¹² cm Hz¹/²/W when thephotodiode is at 200 K. Laboratory measurements support the theoretical analysis.

Two commonly used passivants of Hg₀.₈Cd₀.₂Te, the anodic oxide and ZnS, have been studied by x-ray photoelectron spectroscopy combined with ion sputtering. Chemical depth profiles of anodic oxide films of 360 to 1600 A showed that the oxide composition is constant with depth and independent of oxide thickness. Chemical shifts and line shape analysis of the Cd M₄₅N₄₅N₄₅ Auger transition in the oxide, CdO, Cd(OH)2, and CdTeO₃ demonstrate that CdTeO₃ is the major constituent of the anodic oxide. The oxide composi-tion is interpreted as 44% CdTeO₃, 29% CdTe₂O₅, 17% HgTeO₃, and 10% HgTe2O5. Anodization of HgCdTe depletes the semiconductor of 30% - 40% of its Hg near the interface. The spatial extent of this Hg depletion is a function of oxide thickness for thin oxides (<1000 A) but is a constant (150-200 A) for thick films. No significant change in the Cd concentration is seen. A ZnS film deposited on a chemically etched sample forms a graded interface of a (ZnHgCd)Te alloy. In this case, no Hg depletion is seen. Deposi-tion of ZnS on an anodized substrate in high vacuum leads to a reaction of the Zn with the residual 0₂ in the chamber to form ZnO on the anodic oxide before the ZnS. The ZnO then diffuses throughout the anodic oxide.

A panoramic scanner using Time Delay and Integrating Charge Coupled Device (TDI-CCD) technology has been developed for the U.S. Navy. Real-time signal processing techniques in conjunction with a micro-computer are utilized to accomplish the following: synchronization of ID! charge transport with image scan velocity, electronic on-chip exposure control, signal-to-noise improvement through pixel grouping, removal of dark signature, correction of photo response non-uniformity scene dependent background subtraction and automatic gain control to compensate for scene illumination variations. The system is designed to operate from starlight to daylight scene illumination. Daylight haze penetration is improved by means of low contrast image enhancement which can be achieved through the background subtraction and automatic gain control features. Details of the signal processing and some preliminary imagery are presented.