Volume 2, Number 1

2-1 cover

Early Work at Lincoln Laboratory on GaAs Semiconductor Devices
R.H. Rediker

Thirty years ago, in May of 1959, Lincoln Laboratory reported the first gallium-arsenide diffused diodes ever fabricated at a talk at the Electrochemical Society meeting. Measurements of the diode's switching time clearly showed the usefulness of p-n junction GaAs diodes for high-speed semiconductor components.

Prior diodes in GaAs had been point contacts. The rectification ratios for the diffused diodes were orders of magnitude larger than those for point-contact diodes or for an alloy-junction diode described in another presentation at the same meeting. Thus Lincoln Laboratory was at the forefront in taking GaAs forward from the point-contact age.

Thirty Years of Research and Development in Space Communications at Lincoln Laboratory
W.W. Ward and F.W. Floyd

The goal of Lincoln Laboratory's program in space communications is the development of reliable, affordable systems for military communication. These systems must function dependably—even in the presence of natural or hostile interference.

A broad, total-systems approach has characterized the program since its inception. Lincoln Laboratory has designed, built, tested, and operated communications satellites and their corresponding terminals in real-world environments.

Some of the techniques that have been developed to meet the requirements of military space communications have had or will have important, sometimes unanticipated, applications in the civilian sphere.

Using Langmuir Probes to Measure Ion Velocities in Pyrotechnic Plasmas
D.L. Cohen

The very fluctuations that often complicate measurements of plasma parameters can be used to measure ion velocities in pyrotechnic plasmas. When burned in a vacuum, many pyrotechnic devices will form exhaust plumes that are hot enough to contain small quantities of singly ionized atoms. This pyrotechnic plasma can be maintained in the laboratory for only a relatively short period of time because the chamber fills within several seconds, eliminating the desired low-pressure environment. Furthermore, the fluctuating intensity of the burning pyrotechnic causes correspondingly large fluctuations in the plasma. These fluctuations make dependable determinations of plasma parameter values extremely difficult. Under these conditions an old standby of plasma physics—the Langmuir electrical probe—can provide surprisingly reliable estimates of the average ion velocity in the pyrotechnic exhaust plume. The variation in the pyrotechnic's rate of combustion creates concentrations of ions. These concentrations, or ion clumps, are tracked as they escape the reaction chamber; by measuring the time it takes one of these clumps to traverse a known distance, the ion velocity within the plume can be estimated.

A Compact Programmable Array Processor
F.E. Hall and A.G. Rocco, Jr.

An array processor is a processor optimized to operate on arrays or vectors of data. Typically, a similar operation is performed repetitively on many inputs. Taking advantage of the repetitive nature of this processing, array processors can achieve higher throughput by performing operations simultaneously. This technique is known as parallel computation. Array processors also frequently use a form of parallel computation called pipelining, where an operation is divided into smaller steps, and the steps are performed simultaneously. For example, if we are building a large number of houses, instead of having one crew we can have three; while the first crew digs the cellar and pours the foundation, the next puts up the walls and roof, and the last does the finish work. When one crew finishes, it moves on to the next house site, and the next crew begins work. We could also increase the parallelism by having another set of three crews working on another group of houses at the same time. While parallelism and pipelining increase the performance of array processors, these techniques also complicate software development. We have developed an array processor that simplifies code generation by decoupling the processes of computation and memory address generation. The processor is completely programmable, as opposed to processors that implement an algorithm in dedicated hardware. This is a general-purpose array processor, and can be used for a variety of real-time signal-processing applications.

A Generic Architecture for Wafer-Scale Neuromorphic Systems
J.I. RaffeL J.R. Mann, R. Berger, A.M. Soares, and S. Gilbert

Neural-network architectures emulate the brain by using a parallel, highly interconnected system of simple processing units to solve problems. Systems based on neural networks offer promising solutions to complex problems in recognition, control, and robotics. However, the massive parallelism and high fan-out of neural networks impose enormous interconnection requirements on the integrated-circuit designs that implement them. Wafer-scale integration interconnects many circuits on a single wafer, thus eliminating wire bonds, package pins, and external printed-circuit wiring. A generic wafer-scale device for neural networks, which uses multiplying digital-to-analog converters for programmable synapses and operational amplifiers for summing nodes, has been developed. After each wafer is fabricated, laser cuts and links maybe used to define the network connectivity and provide defect avoidance for yield improvement.

Mass-Transported GaInAsP/InP Lasers
Z.-L. Liau and J.N. Walpole

A new microfabrication technique, mass transport, has made possible the development of a new breed of semiconductor laser devices. Mass-transported buried-heterostructure GaInAsP lasers are ideally suited for application in optical-fiber communication systems. In fact, the technology for mass-transported buried-heterostructure lasers has been transferred to a commercial fabricator, and these lasers are now being incorporated in optical-fiber communication systems. Mass transport has also led to the first practical surface-emitting diode lasers and monolithic two-dimensional laser arrays. These devices are potentially useful for applications in communications, sensors, optical signal processing, optical computing, and efficient high-power lasers.

Design of a Real-Time Imaging and Discrimination System
K.R. Roth, M.E. Austin, D.J. Frediani, C.H. Knittel, and A.V, Mrstik

A novel architecture achieves an extraordinary operating speed by linking conventional computing devices—general-purpose computers, high-speed bulk storage, and array processors—in a parallel, distributed network. The network performs calculations at a rate of 450 million operations per second and transfers data at a rate of 160 MB/s, which satisfies the computation-intense requirements of automatic real-time target imaging, discrimination, and classification.

Adaptive-Optics Compensation through a High-Gain Raman Amplifier
C. Higgs, E.D. Ariel, RE. Player, and L.C. Bradley, III

The applicability of Raman amplification to the problem of high-power laser propagation through a turbulent atmosphere was investigated in a two-part experimental program. The goal of the program was to determine the extent of phasefront preservation possible in a high-gain Raman amplifier. The utility of an unconventional approach to adaptive-optics compensation, known as "Raman look-through," was examined. The Raman look-through technique, which combines conventional adaptive optics using a deformable mirror and wavefront sensor with the nonlinear process of stimulated Raman scattering, performs a wavefront correction on the low-power input to a Raman amplifier, and then amplifies the corrected beam.

The first part of the experimental program determined the degree of phasefront preservation in a high-gain Raman amplifier. The phasefront of the input beam to a Raman amplifier was deliberately aberrated, and the phasefront of the amplified output beam was measured and compared to the input beam. The comparisons indicate that, under the proper conditions, a high degree of phasefront preservation can be achieved.

The second part of the experimental program studied adaptive-optics compensation through a Raman amplifier. Atmospheric turbulence was simulated by specially fabricated quartz plates that produced turbulence-like phasefront distortion on a transmitted beacon. The beacon phasefront was measured by a 69-channel wavefront sensor that commanded a deformable mirror to impose the conjugate phasefront on the Stokes seed to a high-gain large-Fresnel-number Raman amplifier. After amplification, the output Stokes beam retraced the path of the beacon through the simulated turbulence. Measurements of the Stokes beam quality indicated a dramatic improvement with near-diffraction-limited performance. The results of both experiments are in good agreement with theory.

top of page