Patent Application
20100054658
The present invention relates to the field of photonic communications devices and more particularly to a system to allow testing of devices for the transfer of data between a optical bus and electrical components having different clock speeds.
Optical busses operate at high bus speeds, unmatched by even the most advances electronic components. The pairing of such electrical devices with optical devices can lead to latency and conflicts. As optical busses evolve to operate at data rates beyond the capability of current characterization equipment, there is a need to develop test and evaluation methods that allow accurate characterization while decoupling the characterization from electronic test equipment and methods that will induced their own lag and latency to the measurements.
One embodiment of the present invention provides a system for interface between an optical bus and electronic component, the system comprising An optical transfer array for receiving and transmitting data to and from the electronic component; an optical decode/encode matrix reading data transmitted from the electronic component uploaded to the transfer array, and writing control data to switches disposed within the transfer array and providing a plurality of addresses for optical guides in the optical bus.
Another embodiment of the present invention provides such a system wherein the transfer array further comprises a plurality of optical latches controlled by the switches, the optical latches being configured to store the data until it can be retrieved by the optical bus.
A further embodiment of the present invention provides such a system wherein the optical bus is a serial bus.
Yet another embodiment of the present invention provides such a system wherein the optical bus is a parallel bus.
A yet further embodiment of the present invention provides such a system wherein the latches within the plurality of optical latches comprise first and second interferometers.
One embodiment of the present invention provides a system for the transmission of data between an optical bus and an electronic component at a speed independent from a clock speed of the electrical component; the system comprising an optical data storage component communicating with both the optical bus and the electrical component; the optical data storage component being configured to hold data transmitted on the optical bus until the electrical component is available.
Another embodiment of the present invention provides such a system, wherein the storage component comprises an optical latch.
A further embodiment of the present invention provides such a system wherein the optical storage component provides a dual cavity.
Yet another embodiment of the present invention provides such a system wherein the system is provided with high index contrast.
A yet further embodiment of the present invention provides such a system wherein the system has low loss.
The features and advantages described herein are not all-inclusive and, in particular, many additional features and advantages will be apparent to one of ordinary skill in the art in view of the drawings, specification, and claims. Moreover, it should be noted that the language used in the specification has been principally selected for readability and instructional purposes, and not to limit the scope of the inventive subject matter.
One embodiment of the present invention provides a bi-rate adaptive optical transfer solution enabled by a series of complex optical components based on the combination of close coupled optical gain materials and existing photonic structures. These integrated optical components will allow full interface with the high speed optical bus bypassing the power and frequency limitations imposed by the application of silicon modulators, germanium detectors, and their related control circuits.
In one embodiment of the present invention an integrated optical gain material is used to realize an all optical bus architecture. These optical gain structures when combined within existing EPIC optoelectronic processing environment will provide us with the building blocks to develop the more complex components necessary to build out the full architecture.
The basic architecture of one embodiment of the present invention at the cell level is illustrated in
This interface is further illustrated in
In such an embodiment information is moved to and from the electronic circuitry without introducing latency to the operation of the optical bus while at the same time providing a method of simplistic bus arbitration in the case of wavelength division multiplexing (WDM) operation of the busses. Such an embodiment can be extended to allow simultaneous encoding of multiple data words as either multiple wavelengths on a single wavelength or multiple wavelengths on parallel waveguides as shown in
As illustrated in
Because modern military systems require ever increasing data throughput, the initial demonstration vehicle for this technology is a digital signal processor (DSP) architecture. The huge volumes of data necessitate implementation of a large number of parallel cores. Because the DSP processing cores are identical, development cost and risk is minimized and focus can be placed on optimization of the execution core itself and with its interfaces to optical clock distribution and busses.
A Bi-rate optically Addressed Interface, such as that of one embodiment of the present invention, will allow development of photonic communications networks that will enable multi-processors with hundreds to thousands of compute cores. Our proposed technology will allow scalling from multi to mega core architectures.
One embodiment of the present invention provides for photonics employed for architectural elements with high fan out used for processor control such as clocks and reset lines, and also for wide data, address, and control busses where photonics result in significantly less lines and interconnections. Photonic signals do not experience either the performance limitations associated with R-C time constants or increasing power dissipation proportional to increasing capacitance load that occur in metal interconnect. These limitations have already been experienced with current processors such as the Intel Pentium where multiple pipeline clock cycles are required to move from one side of the chip to the other. Similarly, micro-architecture of the TRIPS polymorphic processor is designed to eliminate global routing by making connections only between adjacent tiles because of wiring delay constraints. Optical busses will eliminate these constraints simplifying processor architecture. These connections interface with all electronic processing and memory macros to maintain optimal density and power dissipation.
One embodiment of the present invention provides improved simplicity of the coding. The more complex the programming requirements for any electronic component the greater the number of on chip operations to implement it. Programming is simplified for a multi-processor, because the effects of core relative placement vs. latency and throughput no longer need to be considered. Code designed for a small number of tightly packed cores is able to employ the same approach as a much larger number of cores spread across greater distances. Our proposed architecture enables us to optically control the addressing of locations in a similar manor to an electronic bus making our implementation transparent to the user/programmer. Because the changes to an existing digital signal processor architecture will occur at the physical layer, there is absolutely no adverse effect to supporting existing software or coding methods. Any existing infrastructure would be maintained, and all-electrical and hybrid hardware could be used interchangeably both for software development and operation as needed. The constraints currently associated with massively multiple core processors due to floor planning and wiring capacitance delays simply would not be incurred.
One embodiment of the present invention provides a number of benefits to the operational interactions with the individual electronic cell. Core-to-core communication in existing electrically-based multi-core architectures is already constrained by the effects of distance and R-C time constants. While lithography may allow increasing numbers of cores in a given silicon area, efficiency of throughput is constrained by the ability of a given core to communicate with others at full performance due to these limits in semiconductor physics. Use of optical bussing between cores eliminates the constraints of electrical characteristics and opens the option of universal connections. Although one embodiment of the present invention leans towards Digital Signal Processing (DSP) data processing, general processors employing multiple cores will benefit from this as well.
One over riding factor in any bus architecture are the effects of information latency and any latency in the distribution of information on the overall ability to accurately transmit data in a timely fashion. Latency associated with non-adjacent cores will be eliminated by this architecture, freeing up the ability to code software without regard to location of the execution unit on the die. Externally, latency associated with main memory accesses will be mitigated by optical chip-to-chip implementations. This allows simplification of the existing complex infrastructure for multiple level cache memories and increases throughput due to decreased latency in addressing and data transfer.
Alternatively, because optical power dissipation is not a function of frequency, optical clocking, bussing, and switching will allow significant extension of the practical limits of power-performance. In space-based systems, where the power-performance limitations are considerably more constrained due to limitations to conduction cooling, constrained generation of satellite power, and radiation effects, introduction of photonic elements will mitigate and possibly remove a barrier that has essentially already been reached.
As illustrated in
In one embodiment of the present invention, the bi-rate adaptive interface of one embodiment of the present invention was first demonstrated in optical fiber, but was then integrated into known environment as illustrated in
One embodiment of the present invention will employ two basic configurations of existing components, including micro-resonators and integrated gain materials: a cascaded configuration to enable the basic optical logic functions and a more complex loop/mirror configuration
Both of these configurations are employed in a complementary fashion to form the basic components of the optical data bussing scheme configured according to one embodiment f the present invention.
In one embodiment of the present invention, a nonlinear optical mirror is provided, as shown in
Thus the net effect of such a configuration is reflection of the input pulse from the nonlinear optical mirror. This cancellation is assured in a Sagnac interferometer as the pulses transit identical paths and thus obtain equal phase shifts. By making the coupler unbalanced (70:30) instead of 50:50, the pulses in each direction have unequal intensities.
The nonlinear element then produces unequal phase shifts due to the unequal intensities. They then do not destructively interfere at the output, and the input pulse switches, or couples, to the output path. This intensity-dependent optical switching (or comparison) is shown in the graph is based upon Self Phase Modulation (SPM). Should a different pulse enter the control arm, the nonlinear phase shift is caused by Cross Phase Modulation (XPM).
Because the control pulse enters the loop in just one direction, it allows the coupler to be returned to 50:50 coupling, and the loop will switch between 0 and 100% of the input when the sum of the input plus the control intensity exceeds the loop threshold (differential phase shift of π). This provides a settable/changeable threshold for the optical comparator.
One embodiment of the present invention uses a single optical amplifier body of the nonlinear optical mirror. These small high Q, high finesse structures allow us to multiply the nonlinearity by the finesse of the amplifying optical cavity and control the optical properties of the transiting light with very low intensity input pulses which is critical for an efficient low power all optical switching mechanism.
In one embodiment of the present invention, two nonlinear interferometers create an optical R-S flip-flop or optical latch as illustrated in
This optical latch is triggered on by the set pulse. After storage later it is triggered off by a reset pulse which allows the output from the much faster optical bus to be brought into a cell without slowing it down. By adding a second latch the gating of information to the bus can be controlled at a much greater rate of operation than could be achieved with the electronic component alone. A similar configuration will allow us to store and rapidly gate information to the optical bus. In this configuration the electronic component writes a state to the latch. And once the bus comes active the information is rapidly clocked into the optical bus through the second latch. Slightly simpler output architecture can be achieved by running the output of the latch through our optically addressed bus switch. This can be further extended by combining latches to achieve a serial to parallel converter to rapidly burst in data serially at each wavelength.
Nonlinear interferometers using nonlinear phase shifters have a long development history, and a lot of application names such as NOLM, NALM, TOAD, FROG, UNI, FUNI and on and on. They are used for multiplexers/demultiplexers, switches, add/drop devices, correlators, comparators, optical computing gates, and many more applications. While they have been demonstrated to successfully meet a myriad of application requirements, they have remained a laboratory demonstration device primarily due to the high peak powers needed to demonstrate the performance, and because of environmental sensitivity considerations. There has however been a steady progression in increasing the effective nonlinearity of fibers, and newer alternative, or equivalent nonlinearity devices. Table 1 shows the advances in fiber and device nonlinearity. Devices have shrunk from using 6-10 Km of fiber to hundreds of meters of fiber. New fibers (Bi doped and photonic xtal fiber) will further reduce these lengths. Engineered Nonlinearity Materials (ENM) and their horizon cousin three-dimensional nonlinear photonic bandgap structure devices, provide dramatic increases in nonlinearity (albeit at much shorter lengths—mm's, cm's).
These increases in nonlinearity enhance the obtainable sensitivity; while simultaneously shrinking the size; making the environmental issues easier; and PIC integration possible.
In one embodiment of the present invention, the nonlinearity and the gain media can be combined as shown in
The extension of the original concept that we developed for the new for optical latches (Flip-flops)
In any dual cavity gain arrangement there is a high sensitivity of the modes within the individual cavities to the introduction of an injected signal. The simpler of the two configurations is a continuation of the drive to the simplest possible configuration. It is possible in the above described environment to design two rings having a fixed phase relationship as shown in
The time based action of the proposed coupled optical structures is driven by the fact that the orientation and design of the two structures will force one to introduce a perturbation in the optical mode of the other, The only way to destabilize the configuration is inject a optical signal into the dominant cavity, disrupting its modal structure, allowing the modal formation in the second cavity to occur. When the second cavity is stable, the signal from this free running cavity will inject into the other, preventing a stable mode from forming within it until the modal structure within it is perturbed allowing the other to once again reform the dominant modal structure. Tuning of the coupling cavity between the gain elements allows control of the phase offset.
The optical latch of such an embodiment is triggered on by the set pulse. Much later it is triggered off by a reset pulse which allows the output from the much faster optical bus to be brought into a cell without slowing it down. By adding a second Latch control the gating of information to the bus can be controlled at a much greater rate of operation than could be achieved with the electronic component alone while a similar configuration will allow storage and rapidly gating of information to the optical bus.
The foregoing description of the embodiments of the invention has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise form disclosed. Many modifications and variations are possible in light of this disclosure. It is intended that the scope of the invention be limited not by this detailed description, but rather by the claims appended hereto.
