A review of various photodetector device types, detector array architectures and applications of acousto-optic (A-0) signal processing is presented. The primary design considerations for obtaining high dynamic range, high sensitivity, low crosstalk photodetector arrays are also discussed. The current status and future needs for detector arrays for A-0 signal processing are presented. A review of several novel detector structures and concepts that have potential for future A-0 optical signal processing use is described.

The design and operation of acousto-optic (AU) Bragg cells at center frequencies in excess of 3 GHz are considered. The high frequencies involved severely limit the choice of AO material and mode due to acoustic attenuation. In addition, transducer design becomes increasingly difficult. Experimental results for Bragg cells operating at center frequencies up to 7 GHz are presented.

Thallium arsenic sulfide, T2,3AsS4, is one of a class of sulfosalt crystals which have been recognized as being potentially important for acousto-optic applications. This bi-axial crystal has a very high figure of merit, M₂ = 525, and an anomalously slow shear acoustic velocity, leading to both isotropic as well as anisotropic Bragg cells with high performance capabilities. We have built and tested a number of TASL cells, characterized their performance, and determined their limitations.

Multi-channel Bragg cells can provide over one hundred independently addressable channels for spatial light modulator applications. This paper addresses some of the design considerations for design and fabrication of these devices. An acoustic diffraction model is used to analyze some prototype cases. Device characteristics for several Bragg cells produced are discussed.

Improved spatial light modulators (SLMs) are critical to the success of optical processors. One SLM which shows promise of fulfilling several roles in both analog and digital optical computers is the PRIZ. PRIZ is a Soviet acronym meaning image transformer. Two versions of the device have been reported. Only the version without dielectric layers is addressed in this paper. This version will be referred to as the conducting PRIZ. The version with dielectric layers will be referred to as the insulating PRIZ.

The recent availability of inexpensive, video-addressable liquid crystal displays has generated interest in the use of such devices for optical processing applications. This paper will present some pertinent performance characteristics for a device of this type, the Radio Shack (Tandy Corp.) Pocketvision. Specific characteristics to be considered are screen transmission, linearity, modulation depth, modulation transfer function (MTF), temporal response and optical flatness. Examples of device performance as a coherent and incoherent processing element will be briefly considered.

A twisted nematic liquid crystal cell with two transparent conductive substrates each of which has two metal terminals was studied. It was found that a circular or elliptical light-transmitting area (bright area) was formed in the, cell window by applying four kinds of voltage signals to each terminal. The light-transmitting area was varied by changing the effective voltage, and the cell behavior was consistent with theoretical results for the voltage distribution in the cell window.

The performance of several acousto-optic processors of digital accuracy is examined from the viewpoints of system efficiency and multiplication speed. Performance figures for both bit-serial and bit-parallel processors are calculated and compared with those of their electronic counterparts. Such a comparison shows that the acousto-optic systems do not offer any significant advantage.

An optical-hybrid matrix processor is presented and compared in its speed with a digital electronic processor. Optical-hybrid matrix processors are shown to be far more superior in their speed in solving systems of linear equations. This advantage in speed increases with the increase of the matrix size. The problem of the convergence of the solution using the optical-hybrid is investigated. It is found that even with using elctro-optical systems with an error as high as 5% in the I/O devices, convergence was achieved for matrices with condition numbers as high as 150. Some means of improving the condition number of a matrix are also intoduced.

A space integrating optical linear algebra processor is described and laboratory performance of the system in the solution of nonlinear matrix equations for optimal control are presented. A new matrix partitioning method is described and the accuracy of the analog implementation of this processor is emphasized. This same architecture is capable of high accuracy performance. Different performance measures and their suitability as criteria for performance are also noted and discussed.

The modified signed-digit (MSD) number representation offers full parallel, carry-free addition. A MSD adder has been described by the authors. This paper describes how the adder can be used in a tree structure to implement an optical multiply algorithm. Three different optical schemes, involving position, polarization, and intensity encoding, are proposed for realizing the trinary logic system. When configured in the generic multiplier architecture, these schemes yield the combinatorial logic necessary to carry out the multiplication algorithm. The optical systems are essentially three dimensional arrangements composed of modular units. Of course, this modularity is important for design considerations, while the parallelism and noninterfering communication channels of optical systems are important from the standpoint of reduced complexity. The authors have also designed electronic hardware to demonstrate and model the combinatorial logic required to carry out the algorithm. The electronic and proposed optical systems will be compared in terms of complexity and speed.

The control operator method is reviewed for use in the design of a general purpose computer. The control operator digital optical computer (CODOC) concept appears to provide a suitable basis for beginning the systems engineering of an optical supercomputer.

A high convergence modified fixed increment algorithm capable of adaptively computing linear or polynomial discriminant functions in two-class pattern recognition applications is proposed. A purely digital hybrid optical/electronic implementation of the adaptive algorithm is presented. The proposed adaptive gate is capable of handling either digital or analog inputs for situations involving both deterministic and nondeterministic logic.

Most of the compute intensive SDI problem solving processors, (such as the adaptive deformable mirror processor used to compensate for atmospheric turbulence, the battle management processor(s) for ballistic missile defense, signal processor(s) for radar range/Doppler ambiguity function calculation, and many others), rely on a common set of algorithms found in numerical matrix algerbra. Typically, all of these problems are broken up into a set of linear equations where it is the processors task to solve this set. Algorithmic solutions range from the extensive use of the Fast Fourier Transform to the robust Singular Value Decomposition method. Over the past several years considerable research has been focused on the use of systolic arrays, which, when configured correctly, will process these algorithms at extremely high speeds and with great algorithmic efficiency.

The general problem is as follows: Given a number of binary optical inputs, photodetectors thresholded to provide binary outputs, and a desired functional relationship (memory-free) between the inputs and outputs, find an optimum assembly of passive optical elements that implements the relationship. Here the passive optical elements might be diffracting, shadowing, reflecting, or refracting structures of various numbers, locations, sizes, and shapes; the binary optical inputs might be from a spatial light modulator or an array of laser diodes; the desired relationship might be a combinational logic function with a high proportion of don't-cares (input values or patterns for which the output may be either binary value); and optimality might be specified in terms of maximum fabrication and operating tolerances or minimum complexity. Several aspects of this general problem are considered in the context of optical real-weight and complex-weight threshold logic and optical binary pattern classification.

An optical crossbar interconnected signal processor, proposed previously, is reviewed. The ability to implement difficult-to-parallelize algorithms on this processor with reasonable efficiency is illustrated using optimal least square filtering. An inverse filter for removing the effects of a ringing source from reflected data is used as an example. A large number of Toeplitz matrix equations must be solved to determine the optimum inverse filter length and phase. Efficient parallel computation is required. Equations for Levinson's algorithm are provided and an implementation on the optical system is suggested. Performance estimates suggest that interleaving may he used to improve efficiency from 5% to approximately 30%.

Dynamic range is a key performance parameter in spectrum analyzers. The dynamic range of a Bragg cell spectrum analyzer is generally limited by the dynamic range of the self-scanned photodetector arrays. During the past four years, advances in interferometric techniques have overcome this basic limitation demonstrating spectrum analyzers with nearly twice the dynamic range (measured in decibels) as their earlier counterparts. This paper will describe a system based on the interferometric architecture which is also small and compact in size.

This paper reports recent progress of wideband interferometric acousto-optic spectrum analyzers. A novel electronic cancellation technique was demonstrated that effectively reduced the spurious signal caused by the broadband RF references. Detailed description of the interferometric spectrum analyzer hardware was presented. The system demonstrated a dynamic range of 55 dB above tangential sensitivity.

The purpose of this study was to investigate a new heterodyne AO spectrum analyzer configuration and provide preliminary results for possible applications in real-time spectrum analysis. This spectrum analyzer was unique in that it used an interferometric configuration to generate direct channelized RF outputs at the input frequencies. Acousto-optic spectrum analyzers use the acousto-optic interaction to create a Fourier plane of light in space. Simultaneous spectrum analysis of multiple signals is possible in real-time in a configuration that continuously receives all input signals due to the channelization of the acousto-optic cell. In a nonheterodyne configuration, the frequency is read as positions in spacer. The interferometric spectrum analyzer floods the frequency plane with coherent light to recover the phase information which is modulated onto the diffracted beam by the Bragg cell. Frequency is read spatially but the signal output by the spectrum analyzer can be retransmitted for phase analysis, relative power readings, and other purposes as well. In addition, and perhaps most importantly, the dynamic range is increased by the heterodyne structure.

The architecture, operation, and performance of a new integrated acoustic RF channelizer are described. The all-acoustic device uses a phased array piezoelectric transducer to generate frequency dispersed bulk acoustic waves over a 500 MHz bandwidth. The waves are focused onto a 25 channel receiver electrode array by a cylindrical reflector formed on the opposing surface of the device. This channelizer has the potential of providing higher dynamic range than competing technologies.

The principle of Fourier transformation by wave diffraction has been demonstrated using surface acoustic waves. The RF signal to be analyzed drives a phased array of SAW interdigital transducers that acts like a curved diffraction grating to focus and angularly disperse the generated acoustic waves with frequency. An array of output transducers partitions the dispersed signal spectrum into contiguous narrow bands. The approach is novel and has a number of significant features. The device is passive, linear, and bidirectional, and preserves phase information. Sidelobes and spurious signals can be suppressed by amplitude weighting the input transducer elements and using mode selective output transducers. Several experimental devices have been constructed on LiNbO₃. These were evaluated electrically and with a scanning laser probe.

This paper addresses the problems associated with building compact, environmentally hardened acousto-optic signal processing systems for fielded use. An overview of technological improvements which have aided in the design of such systems is given along with the limitations of current devices.

A general analytical model is developed for an adaptive optical filter. This model is used to explain system performance limitations and relate them to specific imperfections of the hardware implementation. Optimization of the system transfer function shows that 30-35 dB of stable gain can be achieved with reasonable hardware and system tolerances.

An interferometric optical system is configured with an active feedback loop as an optical adaptive filter. The closed loop characteristics of the system are found to be sensitive to alignment and aberrations. An open loop diagnostic procedure is applied to the system to develop stability to a level supporting 30 db adaptively formed frequency rejection notches. The time delay in the closed loop signal path is found to be the limiting factor in system performance.

A single channel AO cell having a large time bandwidth product can store ten times the image area required for a single calculation of a 9 x 9 neighborhood operator. With proper data sequencing this storage capability of the AO cell can result in nearly one filtered output data point for every input pixel from the image. The image enters the AO cell as a series of columns which are 90 pixels long rather than the 9 pixels required by the desired 9 x 9 convolution filter. Nine such columns which sit side by side in the image fit in tandem in the AO cell. A mask adjoining the AO cell simultaneously selects the desired nine pixels from each of long columns and multiplies by the (mask) filter function. A pulsed laser source illuminating the AO cell and filter mask is collected on a single detector giving a single neighborhood calculation. Each step of the column data through the AO cell gives a new output data point corresponding to the filter mask stepping down the long columns of the image. The result is efficient data flow in two dimensions, first by inserting one whole new column at a time, and second by making the columns much longer than is required by the filter height. This implementation of a large two dimensional convolution filter requires each data point from the image only 1.1 times, on the average, which is within 10% of the ultimate data flow efficiency.

A general-purpose space, time and frequency-multiplexed acoustooptic correlator for the synchronization and demodulation of two diverse signals is described. The system achieves an infinite range delay search and complex correlations with full processing gain using multiple m mini-correlations". Signals considered include general I and Q waveforms, MSK (minimum shift keying) and PN encoded data, plus frequency hopped data. The resultant processor achieves the necessary synchronization and demodulation requirements for diverse waveforms (with electronic control of the reference signals provided). The case studies include an identification and a communication application. Novel properties of the processor are its complex-correlation and frequency-selectivity. These attributes are described and applications for them discussed.

A variety of acoustooptic signal processing architectures capable of producing triple products in real time are reviewed. As well a status report on the construction of a high bandwidth interferometric triple product processor is given.

The digital Integrating Post-Processor (IPP) described in this paper has been developed to detect and process the outputs of time-integrating optical signal processors. Major system development activities included optimum sensor selection, processing algorithm formulation, and hardware/software design. The IPP consists of a solid state image sensor, a bank of analog-to-digital (A/D) converters, a custom digital processor, a controlling microcomputer and both video and graphics displays. The IPP has been designed to detect optical signal contrasts as low as one in a thousand.

Optimum performance in an acousto-optic processor often requires signal preconditioning. Necessity for fast programmable adaptable wideband notch filters has led to the use of the surface acoustic wave (SAW) transform domain filter which operates by filtering in the frequency domain. We consider analysis of an acousto-optic (AO) system with a SAW filter as a component. DFT-based simulation methods for the AO spectrum analyzer, AO correlator, and SAW filter are examined. Simulation results are compared to system responses for an AO spectrum analyzer. SAW filter effects in an AO correlator are explored, and simulation results presented. The generalized simulation techniques are shown to be applicable to other AO information/signal processors, and may be utilized to examine new architectures.

Coherent time-integrating optical correlators retain phase information. Due to slowly shifting laser phase and mechanical vibration, long-term phase reliability is not easily achieved. Coherent differential phase-shift keying can overcome these problems. An implementation of coherent differential demodulation suitable for use with optical correlators is described. The stochastic nature of the output of the CDD processor is investigated. Experimental results are summarized, and compared with a simulation of the CDD processor to validate the theoretical processor model.

The optical excision processor is useful for rejecting narrowband interference from broadband signals for improved broadband signal detection. Key performance measures of the optical excision processor include filter resolution, filter depth, and spurious free dynamic range. This paper presents the theoretical limits of these key performance measures for the optical excision processor and relates these performance limits to quantified improvements in broadband signal detection.

We discuss the light signals that are coherently emitted by some optical absorbers sub-sequent to excitation by a series of three laser excitation pulses. In some cases, the light signal emitted by the material duplicates the temporal shape of one of the laser in-put pulses. In other cases, the emitted signal represents the convolution or cross-correlation of two of the laser input pulses. In principle, digitally encoded signals of multi-gigahertz bandwidth can be convolved in real time, or, in suitable materials, stored for long periods. Output signals on the order of a few percent as intense as the inputs are possible. We provide experimental demonstrations of these effects, and discuss their implementation in optical signal processing and data storage systems.

The behavior of a binary, spatial light modulator (SLM) in both the input and Fourier planes of a coherent optical correlator system is investigated using both a computer model and a working laboratory correlator. The computer model, based upon the fast Fourier transform, simulates the effect of the SLMs in the system and permits the examination of such parameters as input noise and shifts and their effect upon the theoretical output. The experimental correlator contains a SLM in the input plane operating in the binary amplitude mode and a SLM in the Fourier plane operating in the binary phase-only mode. The output of the system is compared to the model and analyzed. Significant performance results obtained with the computer model are also examined.

In this paper, we present a new hybrid optical/digital processor that computes the geometric moments using the recently introduced Hartley transform (HT). This transform has the attractive property of being real when the signal is real. We prove an important result that all geometric moments of an image can be computed recursively from the various partial derivatives (near origin) of the HT intensity.