Design for an optical random access memory

Applied Optics, 1986

The results of investigations into the feasibility of incorporating optically bistable elements into an optical processor are presented. Two forms of bistable device, etalons of InSb and interference filters containing ZnSe, have been used in the first experimental demonstrations of digital all-optical circuits. It is shown that cw optical bias beams may be used to hold logic gates sufficiently close to their switch point that the available signal gain allows one to realize indefinitely extensible optical logic. The results of such experiments are presented and the implications they have on the field of optical computing are discussed. It is concluded that parallel arrays can give significantly high rates of digital operations.

IEEE/OSA Journal of Lightwave Technology, 2009

This paper describes the fiber optic loop buffer-based switch in which contention is resolved in the time and wavelength domain. In the loop buffer, tunable wavelength converters (TWCs) are placed in place of semiconductor optical amplifiers (SOAs) as in conventional loop buffer-based architectures. The placement of TWCs inside the buffer facilitate simultaneous read/write operation and dynamic re-allocation of wavelengths and improves the switch performance significantly. It is a well known fact that the re-circulating type buffer structure suffers from circulation limit (maximum revolutions that data can take in the buffer) due to the loss and noise accumulation in the switch. This paper presents a mathematical model to obtain a maximum number of allowed circulations of the data in loop buffer-based switch architecture. This model is derived for various configurations (transparent, noisy, and regenerative) of TWC. The detrimental effect of crosstalk and four wave mixing are shown, and the affect of dispersion on the maximum allowed bit rate is discussed. The minimum length of the loop is also evaluated. Finally, the bounded region is shown (bit rate versus number of wavelengths graph) where memory can work efficiently.

ICENCO'2010 - 2010 International Computer Engineering Conference: Expanding Information Society Frontiers, 2011

We experimentally demonstrate the accuracy of an all-optical S-R latch and an optical S-R flip-flop based on hybrid integrated Mach-Zehnder Interferometers, with Semiconductor Optical Amplifier in each arm (MZI-SOA). The performance of both bistable devices will be studied and compared in terms of extinction ratio and switching times.

The concurrent-read concurrent-write (CRCW) model of shared memory for parallel computation is considered unimplementable in practice, due to limitations of electronic technology. But optical technology opens new possibilities that may overcome these limitations. An optical CRCW memory can be designed, in principle, based on the following ideas. First, optical free-space interconnections eliminate the need for shared switches that su er from contention. Second, practically unlimited fan-in on detectors allows for concurrent writing without serialization. And third, retrore ective optical elements that reverse the propagation of beams of light implement concurrent reads without address decoding and sequential service. It should be noted, however, that the required technology is still immature, and much work is needed before a practical design is reached.

IEEE Journal of Selected Topics in Quantum Electronics, 2000

We demonstrate analytical frequency-domain transfer function expressions for an optical random access memory (RAM) cell that employs two SOA-based ON/OFF switches and two coupled SOA-MZI gates forming an optical flip-flop. Our theoretical model relies on first-order perturbation theory approximations applied for the first time to coupled optical switching structures, resulting to an optical RAM cell frequency response that allows for a qualitative and quantitative analysis of optical RAM memory speed and performance characteristics and their dependence on certain RAM cell device parameters. We show that the transfer function of an optical RAM cell and its incorporated flip-flop device exhibits periodic resonance frequencies resembling the behavior of optical ring resonator configurations. Its free spectral range is mainly dictated by the length of the waveguide that enables the coupling of the two SOA-MZI gates, yielding this coupling length as the dominant memory speed determining factor. The obtained results are in close agreement with experimental observations, demonstrating that optimized RAM cell designs with waveguide coupling lengths lower than 5 mm can enable RAM operation at memory speeds well beyond 40 GHz.

An all-optical flip-flop memory with separate set and reset inputs is presented. The flipflop is formed from two coupled polarization switches that are operated by utilizing the nonlinear polarization rotation in semiconductor optical amplifiers. The concept of the system is explained and experimental results are presented, demonstrating that a contrast ratio of over 20 dB between output states and a switching power of less than −3 dBm can be obtained. The all-optical flip-flop can be utilized in all-optical packet switches.

IEEE Photonics Technology Letters, 2000

A multistate optical memory based on serially interconnected lasers is presented. We show that only one of the lasers can lase at a time, thus, the state of the optical memory is determined by the wavelength of the dominant laser. The light from the dominant laser suppresses its neighboring lasers through gain saturation, but still receives amplification by the active element of the suppressed lasers, compensating for coupling losses. This light passes through each of the successive lasers, simultaneously suppressing and being amplified. By this mechanism, all other lasers are suppressed. A five-state optical memory based on this concept is experimentally demonstrated. The contrast ratio between different states is over 30 dB. Dynamic flip-flop operation based on two different all-optical switching methods is also demonstrated.

An all-optical memory with three states is presented. The optical memory is realized from three-coupled identical nonlinear polarization switches. The state of the optical memory is determined by the wavelength of the memory's output light. In each state, only one wavelength is dominant. The concept is explained and experimental results are presented that demonstrate that a contrast ratio of over 20 dB and a switching power of around 8 dBm can be obtained. This concept can be extended for an all-optical memory with a large number of states, which is crucial to realize 1×N all-optical packet switches.

Optics Communications, 1992

The various possibilities of optically interconnecting 16 × 16-data arrays by shuffle patterns are presented and analysed. The pattern analysis is based on evaluating the skew of each particular path of the shuffle interconnections. Geometric quantities of interest (deflection angle, path length of spatial and frequency pattern) are derived. The switch preserving transformation of 2D shuffle interstage patterns into their equivalent 1D shuffles and vice versa, the way a 1D shuffle overlaps into a 2D shuffle in the 3D physical space, is presented. The costs of these different interconnection schemes are evaluated and the results are compared. (The applied cost measure is the total length of the interconnections.) For example, the interconnection of 256 data has the least costs if the arrays are interconnected by a 4D shuffle.

Optoelectronic Interconnects VI, 1999

In this paper we introduce a micro-optical architecture that uses meso-scopic diffractive optical elements ( DOEs) as three dimensional interconnects in a memory system for high level instruction level parallelism (ILP) processors. By using meso-scopic DOEs we can reduce the scale of integration to the VLSI scale, i.e., the micron scale, achieve submicron alignment tolerance, improve reliability due to monolithic integration, facilitate integration into the current manufacturing infrastructure, and offer the ability for higher bandwidth and high interconnect densities. To this the end we are developing the component technologies needed to realize this system. In this paper we present our work in the development of a theoretical and experimental framework for the design and characterization of meso-scopic DOEs, preliminary experimental results of meso-scopic beam splitters, and a large scale demonstration of the ILP memory system.