Experimental free-space optical network for massively parallel computers

Demonstration of a 280 Gbit/s free-space space-division-multiplexing communications link utilizing plane-wave spatial multiplexing

We demonstrate a 280 Gbit/s free-space space-division-multiplexing communications link incorporating a set of independent tilted truncated plane-waves, each generated by a single mode fiber placed at the back-focal plane of a spherical lens. Each of the seven tilted plane-wave channels are encoded with a 40 Gbit/s 16-QAM signal. Our approach comprises two identical linear fiber-arrays placed approximately 5 m apart. As each fiber array is placed at the back-focal-plane of a spherical lens, each fiber array is effectively placed in a conjugate image plane of the other. A channel crosstalk of less than 26 dB is shown, with a bit-error-rate below the FEC threshold of 3.8×10(-3).

Bandwidth optimization of optical data links by use of error-control codes

A design approach to optimizing the bandwidth of optical data links while simultaneously decreasing the bit-error rate is proposed. Mathematical analysis indicates that bandwidth gains by factors of 10-60 with power gains of as much as 8.9 dB are possible. To achieve these performance levels requires several innovations. First, conventional forward error-correcting codes cannot be used because of their excessive hardware cost. A reasonably powerful multidimensional parity-based error-control code is proposed and analyzed. These codes offer excellent error detection and moderate error-correction capabilities. Most importantly, they can operate at the fast clock rates that are required. Second, a hybrid automatic-repeat-request protocol is exploited to correct complex error patterns. In thermal-noise-limited systems this unique combination allows the optical clock rate to be increased significantly, thereby resulting in large bandwidth increases. The proposed design approach can be used in optical data links in which propagation delays are moderate and is applicable to fibers that exploit wavelength-division multiplexing or time-division multiplexing, one-dimensional parallel-fiber ribbons, and two-dimensional optical data links that use free space or guided waves. Several design examples are illustrated.

Free-space optical relay for the interconnection of multimode fibers

We present results from a system that shows that multimode fibers can be used for both the input and the output of a free-space optical system. The system consists of plastic microlenses integrated with plastic optomechanical components that are suitable for low-cost fabrication and assembly. Such a system opens up opportunities to construct large repeaters and switches for multigigabit ethernet applications by integration with two-dimensional arrays of optoelectronic devices. We demonstrate a 2.5-Gbit/s transmission rate by using commercial vertical-cavity surface-emitting lasers coupled to 62.5-microm core fibers. We consider the design constraints and the capabilities of custom optical modules suitable for mass production.

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.

Design of an optical interconnect for photonic backplane applications

A compact alignment-tolerant interconnect has been developed for use within a prototype modulator-based free-space photonic backplane. The interconnect design encompasses several unique features. Microlens arrays are used, and several beams share each microlens by clustering the optical input-output in a small field about the optical axis of each lens. For simplifying the layout, the optical input and output of each smart-pixel array are clustered separately, thereby allowing a Fourier plane patterned-mirror array to be used in the beam-combination optics. This allows a suitable balance between high interconnection densities and reasonable optical relay distances between adjacent boards to be achieved. The primary advantages of this scheme are the simplicity of the optical design and its alignability, making it ideally suited for high-density interconnection applications.

Design, implementation, and characterization of a hybrid optical interconnect for a four-stage free-space optical backplane demonstrator

A four-stage unidirectional ring free-space optical interconnect system was designed, analyzed, implemented, and characterized. The optical system was used within a complementary metal-oxide semiconductor-self-electro-optic-effect-device-based optical backplane demonstrator that was designed to fit into a standard VME chassis. This optical interconnect was a hybrid microlens-macrolens system, in which the microlens relays were arranged in a maximum lens-to-waist configuration to route the optical beams from the optical power supply to the transceiver arrays, while the macrolens optical relays were arranged in a telecentric configuration to route optical signal beams from stage to stage. The following aspects of the optical system design are discussed: the optical parameters for the hybrid optical system, the image mapping of the two-dimensional array of optical beams from stage to stage, the alignment tolerance of the hybrid relay system, and the power budget of the overall optical interconnect. The implementation of the optical system, including the characterization of optical components, subsystem prealignment, and final system assembly, is presented. The two-dimensional array of beams for the stage-to-stage interconnect was adjusted with a rotational error of <0.05 degrees and a lateral offset error of <3.5 mum. The measured throughput is in good agreement with the lower-bound predictions obtained in the theoretical results, with an optical power throughput of -20.2 dB from the fiber input of the optical power supply to the modulator array and -25.5 dB from the fiber input to the detector plane.

Plastic modules for free-space optical interconnects

A combined optoelectronic and optomechanical packaging technique for the construction of snap-together free-space optical interconnect systems is described. The modules integrate relaying and routing functions by use of transparent optical molded plastic, which can achieve sufficient alignment precision that further adjustment is not required during system assembly. Methods to integrate the optoelectronic chips, such as vertical-cavity surface-emitting laser and receiver arrays with these plastic optical modules are described. Other chips can also be integrated to form optoelectronic multichip modules. These modules can also be designed to accommodate coupling to or from optical fiber arrays. A test-bed system to demonstrate the concept was assembled to a lower precision by use of conventional machining techniques.

Specification for a reconfigurable optoelectronic VLSI processor suitable for digital signal processing

A concept for a parallel digital signal processor based on opticalinterconnections and optoelectronic VLSI circuits is presented. Itis shown that the proper combination of optical communication, architecture, and algorithms allows a throughput that outperformspurely electronic solutions. The usefulness of low-level algorithmsfrom the add-and-shift class is emphasized. These algorithms leadto fine-grain, massively parallel on-chip processor architectures withhigh demands for optical off-chip interconnections. A comparativeperformance analysis shows the superiority of a bit-serialarchitecture. This architecture is mapped onto an optoelectronicthree-dimensional circuit, and the necessary optical interconnectionscheme is specified.

Error correction for free-space optical interconnects: space-time resource optimization

We study the joint optimization of time and space resources withinfree-space optical interconnect (FSOI) systems. Both analyticaland simulation results are presented to support this optimization studyfor two different models of FSOI cross-talk noise: diffraction froma rectangular aperture and Gaussian propagation. Under realisticpower and signal-to-noise ratio constraints, optimum designs based onthe Gaussian propagation model achieve a capacity of 2.91 x10(15) bits s(-1) m(-2), while therectangular model offers a smaller capacity of 1.91 x10(13) bits s(-1) m(-2). We alsostudy the use of error-correction codes (ECC) within FSOIsystems. We present optimal Reed-Solomon codes of various length, and their use is shown to facilitate an increase in both spatialdensity and data rate, resulting in FSOI capacity gains in excess of8.2 for the rectangular model and 3.7 for the Gaussian case. Atolerancing study of FSOI systems shows that ECC can provide toleranceto implementational error sources. We find that optimally codedFSOI systems can fail when system errors become large, and we present acompromise solution that results in a balanced design in time, space, and error-correction resources.

Distributed crossbar interconnects with vertical-cavity surface-emitting laser-angle multiplexing and fiber image guides

We report on the implementation of an optical crossbar interconnectconsisting of a centralized free-space beam-steering subsystem, adistributed array of vertical-cavity surface-emitting lasers andphotoreceivers, a fiber image guide, and a large-core polymer fiberarray. The interconnect can in principle handle more than 350cross-bar channels, but our implementation demonstrated only 240channels. Transmissions of 500-Mbit/s per channel weredemonstrated. Approximately 12.7-dB end-to-end optical channelattenuation was measured. Details of component design, packaging, system integration, and testing are presented and discussed.

Design, implementation, and characterization of an optical power supply spot-array generator for a four-stage free-space optical backplane

The design and implementation of a robust, scalable, and modular optical power supply spot-array generator for a modulator-based free-space optical backplane demonstrator is presented. Four arrays of 8 x 4 spots with 6.47-mum radii (at 1/e(2) points) pitched at 125 mum in the vertical direction and 250 mum in the horizontal were required to provide the light for the optical interconnect. Tight system tolerances demanded careful optical design, robust optomechanics, and effective alignment techniques. Issues such as spot-array generation, polarization, power efficiency, and power uniformity are discussed. Characterization results are presented.

Connection cube modules for optical backplanes and fiber networks

A modular free-space optical system, called the connection cube, for connecting arrays of electro-optic transceivers and fiber-array connectors is presented. The connection cube module provides bidirectional data transfer between four processing nodes on a cube face and can be used as a basic building block for optical backplanes and interconnect networks. An experimental system for connecting four processing nodes is presented and used to examine alignment and packaging issues. An analysis of the dimensional requirements and scaling capability for systems based on this module is conducted. This analysis shows that, when the connection cube module is adapted to vertical-cavity surface-emitting-laser-based point-to-point fiber-array links currently under development, it can connect up to 14 processing nodes with an aggregate data transfer capacity of 112 Gbits/s with 19.6-W power consumption.