A prototype 3D optically interconnected neural network

We report the implementation of a prototype three-dimensional (3D) optoelectronic neural network that combines free-space optical interconnects with silicon-VLSI-based optoelectronic circuits. The prototype system consists of a 16-node input, 4-neuron hidden, and a single-neuron output layer, where the denser input-to-hidden-layer connections are optical. The input layer uses PLZT light modulators to generate optical outputs which are distributed over an optoelectronic neural network chip through space-invariant holographic optical interconnects. Optical interconnections provide negligible fan-out delay and allow compact, purely on-chip electronic H-tree type fan-in structure. The small prototype system achieves a measured 8-bit electronic fan-in precision and a calculated maximum speed of 640 million interconnections per second. The system was tested using synaptic weights learned off system and was shown to distinguish any vertical line from any horizontal one in an image of 4×4 pixels. New, more efficient light detector and small-area analog synapse circuits and denser optoelectronic neuron layouts are proposed to scale up the system. A high-speed, feed-forward optoelectronic synapse implementation density of up to 10⁴/cm² seems feasible using new synapse design. A scaling analysis of the system shows that the optically interconnected neural network implementation can provide higher fan-in speed and lower power consumption characteristics than a purely electronic, crossbar-based neural network implementation     

Multi-emitter Si/GexSi1-x heterojunctionbipolar transistor with no base contact and enhanced logic functionality

We demonstrate multi-emitter Si/Ge_xSi_1-x n-p-n heterojunction bipolar transistors (HBT's) which require no base contact for transistor operation. The base current is supplied by the additional emitter contact under reverse bias due to the heavy doping of the emitter-base junction. Large-area HBT test structures exhibit good transistor characteristics, with current gain β≈400 regardless of whether the base current is supplied by a test base electrode or one of the emitter contacts. These devices have enhanced logic functionality because of emitter contact symmetry. Since device fabrication does not require base electrode formation, the number of processing steps can be reduced without significant penalty to HBT performance     

Comparison of mismatch error shaping in multibit oversampledconverters

A number of mismatch error-shaping schemes for oversampled DACs are compared. Simulation results are presented to show that such a comparison should take account of the distribution of the mismatch errors, since the pattern of errors present affects different algorithms in different ways. A scheme for improving the error tolerance of one mismatch shaping architecture is presented     

Short-pulse, eye-safe Nd:YAG laser using cavity-dumping

A short-pulse 1.444-μm laser based on Nd:YAG technology has been demonstrated. The 1.444-μm is eye-safe. With the cavity-dump technique, a pulse of 50 m× and 14 ns was obtained. The beam quality was excellent with an M² of 1.6 by the use of a telescopic resonator. Silicon-window polarizers were used to suppress the 1.06-μm radiation but showed 1.444-μm absorption as well     

Radiation and scattering from ferrite-tuned cavity-backed slotantennas: theory and experiment

A three-dimensional finite-element method hybridized with the spectral/spatial domain method of moments is presented for the analysis of ferrite-tuned cavity-backed slot antennas. The cavity, which is partially filled with magnetized ferrite layers, is flush mounted on an infinite ground plane with possible dielectric or magnetic overlay. The antenna operates primarily in the ultrahigh-frequency band. The finite-element method is used to solve for the electric-field distribution inside the cavity, whereas the spectral-domain approach is used to solve for the exterior region. An asymptotic extraction of the exponential behavior of the Green's function followed by a spatial evaluation of the resulting integral is used to improve computational speed. Radar cross section, input impedance, return loss, gain, and efficiency of ferrite-tuned cavity-backed slots (CBS) are calculated for various biasing conditions. Numerical results are compared with experimental data     

On asymptotic approximation of a solution of a boundary-value problem for a nonlinear nonautonomous equation

On the basis of periodic Ateb functions, in the resonance and nonresonance cases, we construct the asymptotic approximation of one-frequency solutions of a boundary-value problem for a nonlinear nonautonomous equation.     

Linear crosstalk in 4×4 semiconductor optical amplifier gateswitch matrix

We report a comprehensive crosstalk investigation of a packaged InGaAsP/InP 4×4 semiconductor optical amplifier gate switch matrix, experimentally as well as theoretically. For a fully loaded switch with the same wavelength on all four inputs, all possible switching combinations are analyzed, thus yielding realistic crosstalk figures. Coherent and incoherent crosstalk phenomena are identified, and a switch crosstalk less than -40 dB has been measured     

A new erasing method for a single-voltage long-endurance flashmemory

A new method to erase a standard (double-poly, stacked-gate NOR-type) flash cell is proposed. The method, still using the tunneling mechanism to extract electrons from the floating gate, is based on the concept of keeping the electric field constant during the whole erasing operation. The new method has two main advantages with respect to the conventional one: (1) it does not depend on the supply voltage variation and (2) it allows a better reliability in terms of endurance-induced stress. Results have shown that flash device performances are greatly improved in terms of stability and endurance reliability up to one million cycles     

Double-diffusive convection for a heated cylinder submerged in a salt-stratified fluid layer

Double-diffusive convection due to a cylindrical source submerged in a salt-stratified solution is numerically investigated in this study. For proper simulation of the vortex generated around the cylinder, a computational domain with irregular shape is employed. Flow conditions depend strongly on the thermal Rayleigh number, Ra _T , and the buoyancy ratio, R _ρ. There are two types of onset of instability existing in the flow field. Both types are due to either the interaction of the upward temperature gradient and downward salinity gradient or the interaction of the lateral temperature gradient and downward salinity gradient. The onset of layer instability due to plume convection is due to the former, whereas, the onset of layer instability of layers around the cylinder is due to the latter. Both types can be found in the flow field. The transport mechanism of layers at the top of the basic plume belongs to former while that due to basic plume and layer around the cylinder are the latter. The increase in Ra _T reinforces the plume convection and reduces the layer numbers generated around the cylinder for the same buoyancy ratio. For the same Ra _T , the increase of R _ρ suppresses the plume convection but reinforces the layers generated around the cylinder. The profiles of local Nusselt number reflects the heat transfer characteristics of plume convection and layered structure. The profiles of averaged Nusselt number are between the pure conduction and natural convection modes and the variation is due to the evolution of layers. Received on 13 September 1996     

A novel method for elimination of line-current harmonics insingle-stage PFC switching regulators

This paper studies a particular single-stage power-factor-correction (PFC) switching regulator employing a discontinuous-conduction-mode (DCM) boost-input cell and a continuous-current-mode (CCM) forward output cell. Although this single-stage PFC regulator can provide a reasonably high power factor when its PFC stage is operating in discontinuous mode, substantial reduction in line-current harmonics is possible by applying a suitable frequency-modulation scheme. This paper derives a frequency-modulation scheme and proposes a practical implementation using a simple translinear analog circuit. A quantitative analysis on the total harmonic distortion (THD) of the line current when the circuit is subject to a limited range of frequency variations is presented along with some considerations for practical design. Experimental data obtained from a prototype confirms the effectiveness of the proposed frequency-modulation scheme. The proposed analog translinear circuit allows custom integrated circuit implementation, making it a viable low-cost solution to the elimination of line-current harmonics in switching regulators