Optical computer

From Wikipedia, the free encyclopedia

This article or section needs copy editing for proper spelling, grammar, usage, tone, style, and voice. You can help by editing it now. A guide is available, as is general editing help.

An optical computer is a computer that uses light, instead of electricity, to perform computations.

Technically speaking, an optical computer uses bound electrons in isolating crystals instead of free electrons in transistors for computation. Its digital signals are modulated onto a carrier wave in the visible region. No modulator or demodulator exists, because the base band offers only 10 GHz bandwidth whereas the visible band offers 10 THz. It is similar to performing digital computation by a radio.

One fundamental limit is the size. Optical fibres on an integrated optic chip are ten times wider than the traces on an integrated electronics circuit chip. The crystals have the same cross-section as the fibers, but need a length of about 1 mm and so are much larger than a transistor. Therefore signal traveling times will be large.

A more practical limit is the crystal. Current crystals need light with 1 GW/cm² intensity. And as a typical die (in microelectronics) is about 1 cm², and some absorption takes place, this means kilowatts of power consumption, which only allows pulsed operation, but nanotubes may reduce this in the future.

The biggest advantage in the near future is the synergy with optical telecommunication.

It performs its computation with photons or polaritons as opposed to the more traditional electron-based computation. Optical computing is a major branch of the study of photonics and polaritonics. Electronics computations sometimes involve communications via photonic pathways. Popular devices of this class include FDDI interfaces. In order to send the information via photons, electronic signals are converted via lasers and the light guided down the optical fiber.

No true optical computers are declassified or otherwise known to exist. Some devices that are best classified as switches have been tested in the laboratory. Transistors that are composed entirely of optical components are themselves still very new and experimental.

A fully functional computer is composed of many transistors. The number of them required to constitute a computer is arguable, but probably at least 10 and more often 1,000,000 transistors are required to do general computing tasks.

Currently, no true optical computers yet exist. The problems of design seem to stem from eliminating the conversion from photons to electrons and back. This conversion is necessary now because we don't have all-optical versions of all the myriad switching devices required by a computer.

• An interesting property of optical computers, optical pathways- is they can carry many different frequencies of light over each pathway and the light detector(s) can be filtered to respond to each of those frequencies, depending on the flexibly programmed topology used. Very Large arrays (VLA's) (4 megapixels and above) can be fabricated like large optical arrays, each passing, switching or filtering each of the various frequency laser beams.

• Iteration can be accomplished by feedback, as in gate arrays, where the output is fed into different inputs to provide greater programmed logic combinations. Light pathways can exist in many layers of adjacent silicon by total internal light guide reflection as in fiber optics, except reflection of the beams are in many parallel vertical and horizontal lightguide pathways in the bulk silicon substrate, created by AutoCAD -like step and repeat programmed layout wafer fabrication lightguide pathways.

• Crossover switches are used to switch the light beam onto a new light pathway(s), can be accomplished by optical banyan switches, using non-linear optics or MEMS mirrors to steer a light beam onto or off of its intended path. These are used currently in optical switches for fiber optics. A 2000 x 2000 switch can be used for 4 million pathways, with 4 Mpixel CCDs used as the light detector(s) as in digital cameras, to convert the binary(on-off light) back into the electrical from the photonic realm. Silicon dioxide is glass-like and is transparent to lasers. The input(s) is/are a very large array of VCSELs lasers.

• Beam-splitters and mirrors move the light up/down or left/right in the array by silicon being placed at 45 degree angles like these symbols left or up "/" or right or down "\" or straight through "-" or reflecting "|". Periscopes use these same principles, only these are very large microminiature stacked arrays on silicon substrate, using a few more micrometres of depth for the additional array layouts. Putting a combination of these pathways in stacked interconnected multi-layer VLArrays, with banyan switches (to re-program any one pathway onto another) at the output or CCD detector, before being fedback into the optical inputs, allows more programming combinations, and general programming schemes to be employed in a massively parallel optical computer. refs: SAI1992,2006, OAO, SPIE, Photonics Magazine.

• Many different clock cycles are possible, both async and synchronously at the same time by using different wavelengths for each clock on the same lightguide pathway. Combinoral logic can be used with the presence or absence or each clock color. This can give rise to more complex functions. The need for more clocks becomes apparent with massively parallel independent sectored processors.

• Optical Computing has the main advantages of small size/high density, high speed, low heating of junctions and substrate, dynamically reconfigurable, scalable into larger/smaller topologies/networks, well matched for imaging, massively parallel computing capability and artificial intelligence applications—i.e., neural networks of great complexity.

• The future of computing is leaning towards large parallel arrays using photonics, rather than electronics, but will probably, for all practical purposes, be opto-electronic in nature, due to the current realm of electronic computing prevalence of using representative voltages to denote "0" or "1" binary states. Optical computing uses a direct analogy of presence or absence of the recognized signal medium, many laser frequencies on a single optical pathway. Multiplexing many frequencies of laser light onto and de-multiplexing off of an optical pathway are common place in DWDM fiber optics for long haul data transfers between cities at 10 to 40 Gbit/s. Thin films on surfaces can make excellent filters of light or polarization.

Interestingly, modern (normal) electronic computers are taking on significant radio wave properties by themselves. Since the frequency of the system clocks on fast systems has passed the single gigahertz range, circuit designers must consider that any electronic signal varying at such rates will be giving off radio waves at that frequency. This means that a wire in a computer performs the dual function of a conductor of electricity and a waveguide for a gigahertz frequency radio wave.

[edit] See also

• Optical neural network

[edit] External links

• Optical Computer Architectures: The Application of Optical Concepts to Next Generation Computers, Optical Computer Architectures: The Application of Optical Concepts to Next Generation Computers book by Alastair D. McAulay (1999)
• Architectural issues in designing symbolic processors in optics
• "Dr. Alan Huang, head of the Optical Computing Research Department" [1]
• D. Goswami, "Optical Computing", Resonance, June 2003; ibid July 2003. [2], [3]
• K.-H. Brenner, Alan Huang: "Logic and architectures for digital optical computers (A)", J. Opt. Soc. Am., A 3, 62, (1986)
• K.-H. Brenner: "A programmable optical processor based on symbolic substitution", Appl. Opt. 27, No. 9, 1687 - 1691, (1988)