Wavelength-division multiplexing free-space optical interconnect networks for massively parallel processing systems

Wavelength-division multiplexing (WDM) techniques provide many advantages for building optical interconnect networks for massively parallel processing (MPP) systems. A design for a 1024-channel network for MPP systems based on the interconnection-cached network with vertical-cavity surface-emitting laser (VCSEL) arrays with one wavelength is described. We then show how a WDM version with four different wavelengths can increase the channel density. We also show how a WDM system can reduce the fan-in loss by a factor of 4. All the VCSEL's in each array are of the same wavelength, while different arrays use different wavelengths. We describe our experimental WDM subsystem containing four VCSEL arrays, operating at wavelengths of 843, 950, 970, and 980 nm, and three different WDM filters for multiplexing-demultiplexing. We present the operational results of the subsystem at 1 Gbit/s per channel.     

Feasibility study of a scalable optical interconnection network for massively parallel processing systems

The theoretical modeling of a novel topology for scalable optical interconnection networks, called optical multimesh hypercube (OMMH), is developed to predict size, bit rate, bit-error rate, power budget, noise, efficiency, interconnect distance, pixel density, and misalignment sensitivity. The numerical predictions are validated with experimental data from commercially available products to assess the effects of various thermal, system, and geometric parameters on the behavior of the sample model. OMMH is a scalable network architecture that combines positive features of the hypercube (small diameter, regular, symmetric, and fault tolerant) and the mesh (constant node degree and size scalability). The OMMH is implemented by a free-space imaging system incorporated with a space-invariant hologram for the hypercube links and fiber optics to provide the mesh connectivity. The results of this work show that the free-space links can operate at 368 Mbits/s and the fiber-based links at 228 Mbits/s for a bit-error rate of 10(-17) per channel. The predicted system size for 32 nodes in the OMMH is 4.16 mm × 4.16 mm × 3.38 cm. Using 16-bit, bit-parallel transmission per node, the system can operate at a bit rate of up to 5.88 Gbits/s for a size of 1.04 cm × 1.04 cm × 3.38 cm.     

Integrated optoelectronics for optical interconnections and optical processing

Integrated optoelectronics using III–V compound semiconductor technology has so far shown exciting advances for application in optical telecommunication systems. New applications of this technology are in optical interconnections and signal processing systems. The technology is expected to be very effective in solving the wiring limit in data transmission within electronic systems, using the advantages of optical techniques such as high data transmission rate and high parallelism, and thus improve the performance of overall systems. Optical interconnection devices currently being developed aim both at multiplexing vast amounts of data and exhibiting flexible interconnection functions using the advantageous characteristics of light. Future research is expected to explore new techniques such as that for multiplexing and processing data in the wavelength division as well as for integrating functional devices in two-dimensions. Synergetic collaboration among materials and processing, design and fabrication, and packaging areas is extremely important and this will lead to practical optical interconnections and signal processing systems.     

Practical realization of massively parallel fiber-free-space optical interconnects

We propose a novel approach to realizing massively parallel optical interconnects based on commercially available multifiber ribbons with MT-type connectors and custom-designed planar-integrated free-space components. It combines the advantages of fiber optics, that is, a long range and convenient and flexible installation, with those of (planar-integrated) free-space optics, that is, a wide range of implementable functions and a high potential for integration and parallelization. For the interface between fibers and free-space optical systems a low-cost practical solution is presented. It consists of using a metal connector plate that was manufactured on a computer-controlled milling machine. Channel densities are of the order of 100/mm(2) between optoelectronic VLSI chips and the free-space optical systems and 1/mm(2) between the free-space optical systems and MT-type fiber connectors. Experiments in combination with specially designed planar-integrated test systems prove that multiple one-to-one and one-to-many interconnects can be established with not more than 10% uniformity error.     

Experimental free-space optical network for massively parallel computers

A free-space optical interconnection scheme is described for massively parallel processors based on the interconnection-cached network architecture. The optical network operates in a circuit-switching mode. Combined with a packet-switching operation among the circuit-switched optical channels, a high-bandwidth, low-latency network for massively parallel processing results. The design and assembly of a 64-channel experimental prototype is discussed, and operational results are presented.     

Scalable optical hypercube-based interconnection network for massively parallel computing

Two important parameters of a network for massively parallel computers are scalability and modularity. Scalability has two aspects: size and time (or generation). Size scalability refers to the property that the size of the network can be increased with nominal effect on the existing configuration. Also, the increase in size is expected to result in a linear increase in performance. Time scalability implies that the communication capabilities of a network should be large enough to support the evolution of processing elements through generations. A modular network enables the construction of a large network out of many smaller ones. The lack of these two important parameters has limited the use of certain types of interconnection networks in the area of massively parallel computers. We present a new modular optical interconnection network, called an optical multimesh hypercube (OMMH), which is both size and time scalable. The OMMH combines positive features of both the hypercube (small diameter, high connectivity, symmetry, simple routing, and fault tolerance) and the torus (constant node degree and size scalability) networks. Also presented is a three-dimensional optical implementation of the OMMH network. A basic building block of the OMMH network is a hypercube module that is constructed with free-space optics to provide compact and high-density localized hypercube connections. The OMMH network is then constructed by the connection of such basic building blocks with multiwavelength optical fibers to realize torus connections. The proposed implementation methodology is intended to exploit the advantages of both space-invariant free-space and multiwavelength fiber-based optical interconnect technologies. The analysis of the proposed implementation shows that such a network is optically feasible in terms of the physical size and the optical power budget.     

Self-alignment with optical microconnectors for free-space optical interconnections

A self-alignment technique that uses optical microconnectors forthree-dimensional optics in optical computing systems and opticalinterconnections is proposed. The optical microconnector consistsof an optical plug and a socket. On the output plane of an opticalsystem, optical plugs are made of a photosensitive resin exposed tolight through the optical system. Because the correspondingpositions of the optical plugs are critical to the image formed by theoptical system, a detecting device can be aligned and mounted by theconnection of the optical plugs to sockets formed on the surface of thedevice. Optical microconnectors were experimentally fabricated in areflective block optical system. An alignment accuracy of ~20 mum was attained in the experiment.     

Comparison of fully three-dimensional optical, normally conducting, and superconducting interconnections

Several approaches to three-dimensional integration of conventional electronic circuits have been pursued recently. To determine whether the advantages of optical interconnections are negated by these advances, we compare the limitations of fully three-dimensional systems interconnected with optical, normally conducting, repeatered normally conducting, and superconducting interconnections by showing how system-level parameters such as signal delay, bandwidth, and number of computing elements are related. In particular, we show that the duty ratio of pulses transmitted on terminated transmission lines is an important optimization parameter that can be used to trade off signal delay and bandwidth so as to optimize applicable measures of performance or cost, such as minimum message delay in parallel computation.     

Multi-input optical parallel logic processing with the shadow-casting technique

The lensless shadow-casting technique for coded pattern processing usually accommodates two inputs at a time to perform desired logical operations in parallel. A method of binary encoding is proposed that can accommodate multiple input patterns for simultaneous processing. With the proposed multiple-input encoding a carry-look-ahead technique of binary addition is developed that requires fewer processing steps than the conventional ripple-carry method. Experimental results for a few logic-processing operations are included to establish the validity of the proposed technique.     

Space-variant interconnections based on diffractive optical elements for neural networks: architectures and cross-talk reduction

Optical architectures for fully connected and limited-fan-out space-variant weighted interconnections based on diffractive optical elements for fixed-connection multilayer neural networks are investigated and compared in terms of propagation lengths, system volumes, connection densities, and interconnection cross talk. For a small overall system volume the limited-fan-out architecture can accommodate a much larger number of input and output nodes. However, the interconnection cross talk of the limited-fan-out space-variant architecture is relatively high owing to noise from the diffractive-optical-element reconstructions. Therefore a cross-talk reduction technique based on a modified design procedure for diffractive optical elements is proposed. It rearranges the reconstruction pattern of the diffractive optical elements such that less noise lands on each detector region. This technique is verified by the simulation of one layer of an interconnection system with 128 x 128 input nodes, 128 x 128 output nodes, and weighted connections that fan out from each input node to the nearest 5 x 5 array of output nodes. In addition to a significant cross-talk reduction, this technique can reduce the propagation length and system volume.     

Free-space optical interconnections with liquid-crystal microprism arrays

Liquid-crystal microprism arrays are shown to be useful for providing electrically controlled alignment of optical beams and fixed various free-space optical interconnections. They can deflect closely spaced micro-optical beams individually to any position with high transmittance (95%), high deflection angle (~10°), and low voltage (<2.8 V(rms)). Various fixed optical interconnections can be made simply by changes in the voltages applied to the microprism.     

Multiwavelength optical content-addressable parallel processor for high-speed parallel relational database processing

We present a novel, to our knowledge, architecture for parallel database processing called the multiwavelength optical content-addressable parallel processor (MW-OCAPP). The MW-OCAPP is designed to provide efficient parallel data retrieval and processing by means of moving the bulk of database operations from electronics to optics. It combines a parallel model of computation with the many-degrees-of-processing freedom that light provides. The MW-OCAPP uses a polarization and wavelength-encoding scheme to achieve a high level of parallelism. Distinctive features of the proposed architecture include (1) the use of a multiwavelength encoding scheme to enhance processing parallelism, (2) multicomparand word-parallel bit-parallel equality and magnitude comparison with an execution time independent of the data size or the word size, (3) the implementation of a suite of 11 database primitives, and (4) multicomparand two-dimensional data processing. The MW-OCAPP architecture realizes 11 relational database primitives: difference, intersection, union, conditional selection, maximum, minimum, join, product, projection, division, and update. Most of these operations execute in constant time, independent of the data size. We outline the architectural concepts and motivation behind the MW-OCAPP's design and describe the architecture required for implementing the equality and intersection-difference processing cores. Additionally, a physical demonstration of the multiwavelength equality operation is presented, and a performance analysis of the proposed system is provided.     

Tolerance of optical interconnections to misalignment

The fundamental reasons that determine the tolerance of free-space optical interconnect systems to misalignment are considered. By evaluation of the overlap of single-mode optical beams in the presence of misalignment it possible to determine an optimum beam configuration. It is shown that for any level of misalignment there is an optimum beam diameter that maximizes the coupling of light through the system. Many interconnect systems are not single mode throughout, so the analysis is extended to cover multimode systems. It is shown that, in principle, any level of misalignment can be accommodated by use of multimode beams, although at the cost of reduced channel density. It is shown that the presence of misalignment will mean that the number of channels that can be supported by an interconnect reduces with the length of the interconnect. As possible candidates for passively aligned systems, three example optical systems are analyzed by use of the methods developed.     

Electrospinning of three-dimensional nanofibrous tubes with controllable architectures

This paper reports a novel static method to fabricate three-dimensional (3D) fibrous tubes composed of ultrafine electrospun fibers. By using this unique technique, micro and macro single tubes with multiple micropatterns, multiple interconnected tubes, and many tubes with the same or different sizes, shapes, structures, and patterns can be prepared synchronously. Parameters that could influence the order degree of patterned architectures have also been investigated. It is expected that electrospun tubes with controllable patterned architectures and 3D configurations may be attractive in many biomedical and industrial applications.     

High-throughput optical processors based on semiconductor nanostructures for incoherent-light optical analog computers with parallel image processing

Fast-response optical recording media based on semiconductor nanostructures (CdTe, GaAs) have been developed for image recording and processing at a speed of up to 10⁶ cps, which is 2–3 orders of magnitude higher than the speed of well-known media based on liquid crystals (MIS-LC). The new media are characterized by a photosensitivity of 10^?2 W/cm² and a spatial resolution of 5–10 lines/mm. Methods for the readout of images recorded in the nanostructures are developed and high-speed incoherent-light optical processors based on these structures are created. The possibility of using these processors for building optical analog computers and image correlators is demonstrated.     

Volume-consumption comparisons of free-space and guided-wave optical interconnections

We compare volume-consumption characteristics of free-space and guided-wave optical interconnections. System volume consumption is used as a fundamental measure of various point-to-point space-invariant and space-variant interconnections of two-dimensional arrays of N(1/2) x N(1/2) points. We show that, in free-space and space-invariant situations, although volume consumption for macroaperture optics is O(1)(N(3/2)), where O denotes the order, it is only O(2)(N) for microaperture optics. For free-space and space-variant operations only microaperture optics is possible without fundamental power losses. The corresponding minimum volume consumption is O(3)(N(3)). We show that single microaperture-per-channel implementations of either space-invariant or space-variant operations are, in general, more volume efficient than are their two-cascade microaperture-per-channel counterparts. We also show that, for minimizing volume consumption, the optimum relative apertures F#(opt) for space-variant optical elements are, respectively, (5N)(1/2)/4 for a single microaperture-per-channel geometry and (5N)(1/2)/2 for a two-cascade microaperture-per-channel geometry. In guided-wave or fiber interconnect cases our study shows that the volume consumption for space-invariant and space-variant operations is O(4)(N), with O(4) < O(2), and O(5)(N(3/2)), respectively. Thus an important conclusion of the study is that free-space optics is less volume efficient than is guided-wave optics in both space-invariant and space-variant interconnect applications.     

Comprehensive three-dimensional gas chromatography with parallel factor analysis

Development of a comprehensive, three-dimensional gas chromatograph (GC3) instrument is described. The instrument utilizes two six-port diaphragm valves as the interfaces between three, in-series capillary columns housed in a standard Agilent 6890 gas chromatograph fitted with a high data acquisition rate flame ionization detector. The modulation periods for sampling column one by column two and column two by column three are set so that a minimum of three slices (more commonly four or five) are acquired by the subsequent dimension resulting in both comprehensive and quantitative data. A 26-component test mixture and quantitative standards are analyzed using the GC3 instrument. A useful methodology for three-dimensional (3D) data analysis is evaluated, based on the chemometric technique parallel factor analysis (PARAFAC). Since the GC3 instrument produces trilinear data, we are able to use this powerful chemometric technique, which is better known for the analysis of two-dimensional (2D) separations with multichannel detection (e.g., GC x GC-TOFMS) or multiple samples (or replicates) of 2D data. Using PARAFAC, we mathematically separate (deconvolute) the 3D data "volume" for overlapped analytes (i.e., ellipsoids), provided there is sufficient chromatographic resolution in each of the three separation dimensions. Additionally, PARAFAC is applied to quantify analyte standards. For the quantitative analysis, it is demonstrated that PARAFAC may provide a 10-fold improvement in the signal-to-noise ratio relative to a traditional integration method applied to the raw, baseline-corrected data. The GC3 instrument obtains a 3D peak capacity of 3500 at a chromatographic resolution of one in each separation dimension. Furthermore, PARAFAC deconvolution provides a considerable enhancement in the effective 3D peak capacity.