Such state-of-the-art devices as multielement linear laser diode arrays, multichannel acoustooptic modulators, optical relays, and avalanche photodiode arrays, are presently applied to the implementation of a 32-bit supercomputer's general-purpose optical central processing architecture. Shannon's theorem, Morozov's control operator method (in conjunction with combinatorial arithmetic), and DeMorgan's law have been used to design an architecture whose 100 MHz clock renders it fully competitive with emerging planar-semiconductor technology. Attention is given to the architecture's multichannel Bragg cells, thermal design and RF crosstalk considerations, and the first and second anamorphic relay legs.

In the past 15 years, a dozen or so designs have been proposed for optical computers. Of these, maybe one-third of them have actually been built and only a few of those tested. This paper will give an overview of some of the systems that have been built as well as mention some promising early and current designs that have not been built. Emphasis will be given to application oriented designs that have been built and tested. This means, of course, that the bulk of the paper is devoted to optical correlators.

This report is for the individuals who need to know the operating principals, specifications and availability of Spatial Light Modulators (SLM's) used in optical processing. This is to aid in decision making. The units presented here have been used by researchers to demonstrate the principles of optical Computing, optical Correlation for Target Recognition, optical Interconnects for Communications and optical Displays. In just a few cases, Semetex Magneto-optic SLM and Hughes LCLV, the SLM's have actually been used in system products. An SLM selection guide has been included for quick reference.

Here, we introduce a general nonlinearly-recorded matched filter that we term as the Fractional Power Filter (FPF). The Classical Matched Filter (CMF), the Phase-Only Filter (POF) and the Inverse Filter (IF) are special cases of the FPF. It is shown that by varying the degree of nonlinearity used in forming the FPFs, we can trade-off the noise tolerance of the filter against its correlation peak sharpness. It is also shown that the fractional power concept can be extended to distortion-invariant composite filters.

We discuss experimental measurements of both spatial and temporal optical transfer characteristics of an optically addressed spatial light modulator (OASLM). The device incorporates a hydrogenated amorphous silicon (a-Si:H) photodiode and a surface-stabilized ferroelectric liquid crystal (SSFLC) modulator. Although the switching of SSFLCs is essentially binary at the microscopic level, the OASLM is shown to have an analog optical transfer characteristic. This is believed to be the result of both spatial and temporal averaging.

The Navy is currently investigating the possible use of cepstral processing in existing and future fire control subsystems. The cepstruxn is difficult to implement in real time using digital electronics. However, if a real time optical logarithm can be generated, then the optical cepstrum can be implemented in a real time system. A real time optical logarithmic transform has been demonstrated at frame rates of 10 kHz using the thresholding capability of an Amorphous Silicon, Ferroelectric, Liquid Crystal, Spatial Light Modulator (ASFLC-SLM). As such, it is an advance over current systems, which use either photographic film as a thresholding device or a Hughes light valve that operates at frame rates of less than 50 Hz. Using experimental data, the relation between the input and output intensities of the system has been shown to be logarithmic over the input range of three to ten milliwatts. Using characteristics of our halftone screen and ASFLC-SLM, it has been possible to compute theoretical values of the output intensity from this system. These results agree nicely with those obtained experimentally.

Two novel deformable mirror structures have been developed for spatial light modulators: an 'AM torsion beam' and a 'phase-mostly single-quadrant cantilever' beam. Both devices are well-suited to optical correlator input and filtering functions. Which the optical modulation characteristic of the torsion-beam modulator is essential amplitude only, which is well suited for use as the input modulator of the optical correlator, the characteristic of the one-quadrant modulator is a phase-mostly modulation whose amplitude changes are coupled to the phase changes; this renders it operable in the Fourier plane of the optical correlator as the filtering device.

The operational characteristics of two alternative types of deformable mirror device (DMD) for optical signal processing applications, respectively designated the 'cantilever beam-inverted cloverleaf' and the 'flexure-hinged piston-type', have been simulated by full-scale parallel computer models. Attention is given to Fourier plane and correlation plane analysis results for each DMD type's use as an input device and as a phase-modulating filter device. Simulated system responses are evaluated for several correlator configurations, and the parallel algorithms used to both implement the models and perform the Fourier transforms are evaluated to ascertain the speed and effectiveness of the parallelism relative to serial algorithms.

In this paper we consider the problem of detection and classification of signals in the presence of additive Gaussian noise of unknown covariance (AGN/TJC), using higher than second-order statistics (HOS) of the output of a matched filter. Specifically, we apply the HOS-based method developed in [1,2] to phase-only matched filters. The main result of this paper is that the HOSbased statistic is appropriate for use with phase-only matched filter (POMF) outputs. Simulation results are presented which indicate the ability of the matched filter and the POMF, which are augmented with 1105, to detect a 2-d signal at signal-to-noise ratios below which the matched filters alone are incapable of making a detection.

Phase-Only filtering1 has received much attention recently because of its 100% light throughput efficiency and the resulting sharp correlation peaks. However, due to its wide bandwidth, conventional POF is very sensitive to noise in the input image. We have previously proposed that by turning off (i.e., by setting the magnitude to zero) some frequencies, we can improve the output Signal-to-Noise Ratio (SNR). We termed such filter' as Optimal POFs (OPOFs). In this paper, we present an algorithm for designing a support function (which indicates the frequencies of the Phase-Only filters with unit magnitude) to get the sharpest possible output correlation peaks. The rest of this paper is organized as follows. Next section introduces the performance measure used to characterize the sharpness of the correlation peak. An algorithm is developed in section 3 to maximize this figure of merit. Section 4 describes some of the computer experiments that illustrate these ideas and the final section provides our conclusions.

Optical demonstration of phase-only circular harmonic filtering using a binary CGH in a standard Fourier transform correlator is reported here. The experimental results have confirmed the computer simulated results1 in that the phase-only circular harmonic filtering produces smaller side-lobes and more prominent correlation peaks with the target.

Experimental results of an optical binary phase-only correlator using both visual and IR imagery are presented. The inputs to the correlator originate from actual aerial imagery containing aircraft and a variety of distortions. Filters used as a database for the system are derived from models of aircraft. Digital image processing techniques are used on images before being input into the optical correlator to enhance the performance of the system. Both noise removing and segmentation techniques are investigated. Input images to the correlator are displayed on a 128 x 128 magnetooptic spatial light modulator (SLM). Experimental results are presented which show that the system performs well with images which are easily segmented from the background.

Pattern recognition invariant to image rotations of up to 75 deg, using a single filter, has been demonstrated for binary synthetic discriminant function (BSDF) optical filters, suggesting their use in directed graph-arranged data bases which can be rapidly traversed by means of a filter-plane programmable spatial light modulator (SLM). The filter data base is arranged as a tree structure in which the root node filters are invariant to over 60 deg rotation, while the leaves are 5-deg invariant. Results are presented from experiments with BSDFs designed to recognize in-plane-rotated views of a Space Shuttle Orbiter. Using a magnetooptic SLM that is driven at 350 Hz in the filter plane, orientation identifications requiring less than 30 msec have been achieved after sequencing through only 10 BSDFs.

A novel channelized, hybrid time- and space-integrating acoustooptic (AO) spectrum analyzer is described. The architecture consists of two AO cells in a crossed-cell configuration. The first AO cell is a wide bandwidth device that performs space-integrating spectral analysis and channelizes signals according to carrier frequency. The second AO cell, in conjunction with a modulated source, performs time-integrating spectral analysis of the signal envelope using the chirp algorithm. One possible application of the processor is to determine the carrier frequency and pulse repetition frequency (PRF) of received radar signals.

The location of a fixed transmitter may be determined by processing the signals collected by two moving receivers. This process involves the simultaneous determination of the difference in propagation delay from the transmitter to the receivers and the rate of change of delay difference induced by the motion of the receivers. This paper describes a system approach to the real time computation of a cross ambiguity function (CAF) for wideband signals, such as frequency hopping radios, that is designed around a wideband optical processor. The optical processor subsystem generates a wideband CAF (delay, delay rate of change coordinates), as opposed to a narrowband CAF (delay, frequency shift coordinates), includes a narrowband interference suppression capability and high dynamic range. A system functional description is presented along with location performance estimates. The optical processor subsystem concept, a compact implementation approach and estimated performance are also discussed.

A symmetric acoustooptic time-integrating architecture able to compute the cross ambiguity function is described. An application of this new architecture is the generation of range-Doppler images from laser radar returns. This symmetric architecture has several advantages over the asymmetric time-integrating architecture, particularly in lowered optical power requirements and the elimination of background ridges. Specific advantages of the reduced optical power requirement are that (1) laser diodes are more likely to be able to satisfy the corehent optical power requirement, and (2) the risk of optical damage to the Bragg cells is diminished.

A folded optical rf spectrum analyzer is described which is built around the 128 x 128 Frame- Addressed Spatial Light Modulator (FASLM). The input waveform is sampled at 32 MHz and displayed on the FASLM at a 2 KHz frame rate. A time-bandwidth product of approximately 8000 is achieved. A brief description of the device operation and optical performance is included.

This paper describes an acoustooptic signal processor designed to produce real-time spotlight mode synthetic aperture radar (SAR) images for autonomous missile guidance. The processor is designed to maintain a range and azimuth resolution of 2 m. High-quality image formation results from electronic motion compensation applied to each received pulse. The processor controls range sidelobes through a weighting function applied during the correlation process, while azimuth sidelobes are controlled either by weighting or by increasing the corehent integration time.

An optical ambiguity function processor fabricated from off-the-shelf components, which is augmented with preprocessing and postprocessing electronics, is presently demonstrated to process the return echoes of each radar pulse to form the image of the terrain illuminated by a SAR in real time. The reconstruction of the SAR image proceeds by compressing the signal first in azimuth, and then in range, by means of an ambiguity-function generator, which measures the Doppler effect imparted onto the return signal by a target and then measures the range to the target on a pulse-by-pulse basis.

A novel acoustooptic architecture that implements linear least-mean square adaptive filtering and prediction and has the potential for very wide bandwidth signal processing is described and analyzed in this paper. This architecture maintains coherent operation (providing complex weights) and employs a single acoustooptic device, a photodiode, and a laser diode. The application of a multichannel acoustooptic processor to multiple antenna adaptive interference cancellation is then presented.

Multi-channel spatial light modulators are used to excise frequency bands of an RF signal in an optical processing system. An advantage realized in using an optical implementation is the minimal phase distortion introduced into the pass band when a narrow-band notch is implemented. Measured amplitude and phase transfer functions are presented for a 200 MHz bandwidth system with 1 MHz channels.

Domains other than the time or frequency domains arises naturally in coherently illuminated optical systems that produce Fourier transforms. We extend the well known result that N samples are sufficient to represent the information content in the object, image, and Fourier planes to include any Fresnel plane of a coherently illuminated optical system, provided that we use a specified sampling technique.

This paper describes the architecture and performance of a prototype Acousto-Optic RF Signal Acquisition System designed to intercept, automatically identify, and track communication signals in the VHF band. The system covers 28.0 to 92.0 MHz with five manually selectable, dual conversion; 12.8 MHZ bandwidth front ends. An acousto-optic spectrum analyzer (AOSA) implemented using a tellurium dioxide (Te02) Bragg cell is used to channelize the 12.8 MHz pass band into 512 25 KHz channels. Polarization switching is used to suppress optical noise. Excellent isolation and dynamic range are achieved by using a linear array of 512 custom 40/50 micron fiber optic cables to collect the light at the focal plane of the AOSA and route the light to individual photodetectors. The photodetectors are operated in the photovoltaic mode to compress the greater than 60 dB input optical dynamic range into an easily processed electrical signal. The 512 signals are multiplexed and processed as a line in a video image by a customized digital image processing system. The image processor simultaneously analyzes the channelized signal data and produces a classical waterfall display.

Acousto-optic interferometric spectrum analyzers are important for signal detection and signal processing applications due to their dynamic range and their coherent detection capabilities. The inability to fabricate a photodetector which matches the available dynamic range of the optical subassembly often limits the performance of an interferometric spectrum analyzer. The photodetector fabrication can be simplified by designing the interferometric spectrum analyzer in a way that a common IF is present at each pixel location. The conventional approach to generating a constant IF in an interferometric spectrum analyzer is to mix the incoming signal with a reference signal on the photodetector element. When this technique is used to produce the IF, an interferometric spectrum analyzer contains spurious signals in its output with no input signal as a result of the overlap of the reference signals on the photodetectors. An alternate approach to producing the constant IF desired for the photodetectors is discussed in this paper. The described approach does not generate spurious signals when no input signal is present.

Coherent light, in the presence of aberrations, produces different focal plane effects than does incoherent light. These are examined for primary spherical aberration, coma, tilt, and defocus. Also discussed are two incoherent light systems.

Efforts to improve the state-of-the-art in charge-coupled device (CCD) detector arrays for acoustooptic applications have had considerable success. A program for the improvement of detectors for acoustooptic applications sponsored by Harry Diamond Laboratories in conjunction with Loral Fairchild Imaging Sensors, formerly Fairchild Weston Systems, is aimed at increasing high speed CCD detector dynamic range to over 65 dB. Other device specifications such as crosstalk and image lag are being addressed in order to match the increased dynamic range performance in a 1024 element detector. This paper outlines the key device specifications which needed to be addressed, the methods used to achieve improved performance in these areas, the testing considerations, and the results of recent device testing.

Rotation of the incident optical beam in Te02 allows selection of the tangential AO interaction to achieve higher efficiency-bandwidth product. This interaction mode results in improved diffraction efficiency by having a longer interaction length, an increase in the two-tone dynamic range, and allows for the use of linear laser polarization. A detailed analysis shows the distinction between the Z axis incident wave and the rotated incident wave in phase mismatch conditions and the efficiency bandwidth relationship. Two devices were fabricated with the following characteristics: aperture time of 70 microseconds with 30 MHz bandwidth; and aperture time of 50 microseconds with 40 MHz bandwidth. A brief result of the phase mismatch analysis, design of the devices, and measurements of the performance compared against theory are presented.

A technique for the measurement of acousto-optic intermodulation products and the the improved dynamic range performance of a Gaussian weighted phased array slow shear Te02 50 MHz bandwidth Bragg cell are reported. The intermodulation products caused by the nonlinear acousto-optic interaction dominated over those caused by the elastic nonlinearity for most of the 50 μs cell length except near the end. This represents an improvement over the conventional transducer design where the nonlinear elastically generated intermodulation products dominate and limit the device dynamic range. The interferometric measurement system had over 50 dB of dynamic range and could easily display all of the two tone intermodulafion products with a standard RF spectrum analyzer.

InP acoustooptic Bragg cells which are IR-transparent in the 1-10 micron bandpass have a center frequency in the 200-600 MHz range, and a diffraction efficiency of 40-60 percent, on the basis of 1-W RF driving power. These devices are anticipated to be ideal in such applications as fiber-optic modulators, IR scanners, deflectors, and HF mode-lockers. In the course of fabrication, the photoelastic constant p44 has been defined; using other crystallographic configurations, such photoelastic constants as p11 and p12 are expected to emerge.

A new scheme for digital optical computing, utilizing a non-holographic opto-electronic addressable memory (CAM), is discussed. To illustrate the performance of this arithmetic processor, the design of an optical binary carry look-ahead adder (CLA), also, the design of a binary, a logarithmic number (LN) and a residue number (RN) multipliers are presented. Compared to existing opto-electronic approaches, this non-holographic CAM offers a number of practical advantages, such as fast processing speed, ease of optical implementation and alignment. Two spatial input data encoding techniques, an active low and high, are discussed. A multioperation multibit CAM processor is presented. Experimental results for a CLA adders; and a binary, residue and logarithmic number multipliers are also presented.

Error sources in an optical matrix-vector processor are analyzed in terms of their effect on the performance of the algorithms used to solve a set of nonlinear and linear algebraic equations. A direct and an iterative algorithm are used to solve a nonlinear time-dependent case-study from computational fluid dynamics. A simulator which emulates the data flow and number representation of the OLAP is used to studs? these error effects. The ability of each algorithm to tolerate or correct the error sources is quantified. These results are extended to the general case of solving nonlinear and linear algebraic equations on the optical system.

A ground-based adaptive optics imaging telescope system attempts to improve image quality by detecting and correcting for atmospherically induced wavefront aberrations. The required control computations during each cycle will take a finite amount of time. Longer time delays result in larger values of residual wavefront error variance since the atmosphere continues to change during that time. Thus an optical processor may be well-suited for this task. This paper presents a study of the accuracy requirements in a general optical processor that will make it competitive with, or superior to, a conventional digital computer for the adaptive optics application. An optimization of the adaptive optics correction algorithm with respect to an optical processor's degree of accuracy is also briefly discussed.

We show that both from the computational viewpoint and from the photon efficiency viewpoint, digital processors require more information input than analog processors to achieve the same result. Thus, both can accomplish the same tasks (a version of Church's thesis) but only at a previously unnoticed price.

One of the features in neural computing must be the adaptability to changeable environment and to recognize unknown objects. This paper deals with an adaptive optical neural network using Kohonon's self-organizing feature map algorithm for unsupervised learning. A compact optical neural network of 64 neurons using liquid crystal televisions is used for this study. To test the performances of the self-organizing neural network, experimental demonstrations with computer simulations are provided. Effects due to unsupervised learning parameters are analyzed. We have shown that the optical neural network is capable of performing both unsupervised learning and pattern recognition operations simultaneously, by setting two matching scores in the learning algorithm. By using slower learning rate, the construction of the memory matrix becomes topologically more organized. Moreover, by introducing the forbidden regions in the memory space, it would enable the neural network to learn new patterns without erasing the old ones.

An architecture for general-purpose optical neural network processor is presented in which the interconnections and weights are formed by directing coherent beams holographically, thereby making use of the space-bandwidth products of the recording medium for sparsely interconnected networks more efficiently that the commonly used vector-matrix multiplier, since all of the hologram area is in use. An investigation is made of the use of computer-generated holograms recorded on such updatable media as thermoplastic materials, in order to define the interconnections and weights of a neural network processor; attention is given to limits on interconnection densities, diffraction efficiencies, and weighing accuracies possible with such an updatable thin film holographic device.

An all-optical processor for sensor inputs capable of computing the resultant strain and optical control signals for such applications as underwater fiber-optic sensor arrays and smart aerospace structures is presented. Attention is given to computer simulation and experimental results obtained to date, which indicate that neural network architectures can perform the very high speed calculations with the requisite accuracy. The performance thus far obtainable in the available bistable optical gate arrays is the basic performance limitation identified for this all-optical implementation.

A new algorithm using the Householder transform to directly update an inverse data covariance matrix is examined for adaptive phased array radar (APAR) problems This algorithm is shown to operate very efficiently on parallel matrix-vector processors without the need for costly vector outer products. The processing accuracy requirements for this algorithm are shown to be roughly half that of more conventional direct approaches, which nearly doubles the processing speed. We investigate the performance of this algorithm with various jammer scenarios, and incorporate a joint jammer-clutter reduction architecture using a displaced phase center antenna (DPCA).

Abstract not available.

This paper reports the design and test of a high performance wideband heterodyne channelizer with low spurs due to reference-reference mixing. The measured signal to noise density ratios agree with the analytical predictions to within 1 dB in the center channels and 3 dB on the outer channels. Asymmetric optical power splitting, unequal aperture times and dual detector cancellation were utilized to achieve low spurs.

Two acoustooptic spectrum analyzers are described. In both system the detector array is replaced by an acoustooptic Bragg cell driven by a short RE pulse and a single fast photodetector. With the power spectrum analyzer system a time-bandwidth product (TBWP) of > boo, a speed of <5o ,us, and a dynamic range > 4o dB can be achieved. Mixing the diffracted beam with a local oscillator an interferometric system is realized where an increased dynamic range of 7o dB can be achieved with the same speed and TBWP.

Spatial light modulators (SLM) are examined from the viewpoint of miniaturization and optical correlation systems. The structure and behavior of a silicon-chip active-matrix-addressed ferroelectric liquid-crystal SLM is presented. The response of an amorphous-silicon optically addressed nematic liquid crystal SLM is reported.