Professional Documents
Culture Documents
Optical Computers PDF
Optical Computers PDF
GYAN VIHAR
ScHool of ENGINEERING & TEcHNoloGY
A
Seminar Report On
OPTICAL COMPUTERS
Submitted in Partial Fulfilment for The Award of Degree
B.Tech. (Computer Science & Engineering)
By
Rajasthan Technical University, Kota
Session 2009-10
1
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Contents
Overview of Optical computers
1 Components of Optical computers. . . . . . . . . . . . . . . . . . . . . . 9
1.1 Hard Disk
1.2 CPU
1.3 Memory
1.4 Cache Memory
1.5 Main Memory
1.6 Screen
1.7 Power Supply
4 Fibre Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.1 Use of Fibre Optics in Computing
4.2 Why use Fibre Optics
5 An Optical Computer Powered by Germanium Laser. . . . 40
Application
Merits
Drawback
2
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
An optical computer (also called a photonic computer) is a device that uses the
photons of visible light or infrared (IR) beams, rather than electric current, to
perform digital computations. An electric current creates heat in computer systems.
As the processing speed increases, so does the amount of electricity required; this
extra heat is extremely damaging to the hardware. Light, however, creates
insignificant amounts of heat, regardless of how much is used. Thus, the
development of more powerful processing systems becomes possible.
On October 4, 1993, the eminent Soviet physicist Prof. U. Kh. Kopvillem would
have been 70 years old. However, he died prematurely on September 24, 1991.
His research was the foundation of several areas of nonlinear optics, quantum
acoustics, and radioacoustics. The breadth of the subject matter of this issue,
ranging from studies on the role of photon modes in high-temperature
superconductivity to the propagation of ullxashort pulses (of the order of one
period), only partially reflects the wide specmam of the scientific interests of U.
Kh. Kopvillem.
3
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
4
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
For decades, silicon, with its talent for carrying electrons, has been
the mainstay of computing. But for a variety of reasons (see "The Coming Light
Years"), we're rapidly approaching the day when electrons will no longer cut it.
Within 10 years, in fact, silicon will fall to the computer scientist's triple curse:
"It's bulky, it's slow, and it runs too hot." At this point, computers will need a new
architecture, one that depends less on electrons and more on... well...what else?
Computer of 2010
5
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Optics. With the assistance of award-winning firm frogdesign (the geniuses behind
the look of the early Apple and many of today's supercomputers and workstations),
Forbes ASAP has designed and built (virtually, of course) the computer of 2010.
The result is a computer that is far more reliable, cheaper, and more compact
—the entire thing, believe it or not, is about the size of a Frisbee--than the all-
electronic solution. But above all, optoelectronic computing is faster than what's
available today.How fast ? In a decade, we believe, you will be able to buy at your
local computer shop the equivalent of today's supercomputers.
How likely is it that this computer will be built ? Some of its components
are slightly pie-in-the-sky. But many others have already been developed or are
being developed by some of the best scientific minds in the country. Sooner or
later, and probably sooner, an optoelectronic computer will exist .
When we plug our 2010 PC into the wall of our home, our house will
become smart, anticipating our every desire. At work, we'll plug it into our desk,
which will become a gigantic interactive screen. When it communicates wirelessly
with a small mobile device, we'll have a personal digital assistant—on steroids.
6
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
7
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
few optical fibers and films, making the systems more efficient with no
interference, more cost effective, lighter and more compact. Optical components
would not need to have insulators as those needed between electronic components
because they donot experience cross talk. Indeed, multiple frequencies (or different
colors) of light can travel through optical components without interfacing with
each others, allowing photonic devices to process multiple streams of data
simultaneously.
SECURITY
9
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Visible-light and IR beams, unlike electric currents, pass through each other
without interacting. Several (or many) laser beams can be shone so their paths
intersect, but there is no interference among the beams, even when they are
confined essentially to two dimensions. Electric currents must be guided around
each other, and this makes three-dimensional wiring necessary. Thus, an optical
computer, besides being much faster than an electronic one, might also be smaller.
10
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
known .
The third problem sets a lower limit to the size of conventional optical
components and hinders the construction of an optical computer on a molecular
scale. However, the development of molecular optics would reduce the size of such
components by a factor of 500.
The limitation of resolution by the wavelengths of light may be overcome
by the transport of the energy of light instead of the emission and absorption of
light quanta. This corresponds to the use of the alternating current (50 Hz) with a
problematic wavelength of some 6000 km where the electrical energy is handled
on a human scale or even lower.
In analogy to such a transport of electrical energy an energy transfer
between chromophores can replace the absorption and emission of light quanta in
optical signal processing components. The transfer will proceed rapidly if the
distance between the two chromophores lies within the F¨orster radius, that means
between 2 and 3 nm for most combinations of similarly absorbing chromophores.
On the other hand, this F¨orster radius would be the natural lower limit for the size
of complex arrangements of switching components for handling energy transfer
because going below this limit would spread energy over many chromophores
without control; a solution of this limiting problem would be the prerequisite for
the development of optical computers with very high densities of integration.
11
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
• Hard Disk
• CPU
• Memory
• Cache Memory
• Main Memory
• Screen
• Power Supply
12
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Our 2010 CPU will operate on the same principle as today's PCs.
But instead of electronic microprocessors providing the brains and
brawn, our future CPU will have optoelectronic integrated circuits
(chips that use silicon to switch but optics to communicate). This
will give us huge increases in speed and efficiency. Why? Because
the CPU of today spends far too much time waiting around for data
to process. Instantaneous on-chip optical communication, and
memory running as fast as the processor, will guarantee a
continuous stream of data processing within the CPU. With
communication between components no longer bottlenecked by
electronic transmission, we can probably push the clock rate to 100
gigahertz.
Our universal appliance of tomorrow also has a
hexagonal optoelectronic processor surrounded by a ring of fast
cache, so that data for any part of the chip can be fetched from the
closest part of the cache. The result will be computer
performance--or, at any rate, delivery of computational results--
comparable to today'ssupercomputers .
13
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
(3) MEMORY(RAM)
14
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
To build our new fast cache, we'll need to get rid of the
inefficiencies of today's product, which requires the computer to
constantly refresh it, just like short-term memory in humans needs
to be constantly refreshed or it's forgotten. The inefficiencies in
cache are so bad, in fact, that once you know the speed of your
cache you can assume that its real-world performance will be about
a third of that--the missing two-thirds being sacrificed to refresh
cycles.
15
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
16
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Size does matter in our 2010 computer screen. It will either be very
large, literally the desk top of your desktop, or very small, a
monocle you hold up to your eye. For the bigger version, our
computer screen will depend on some kind of photonically excited
liquid crystal, with power requirements significantly lower than
today's monitors. Colors will be vivid and images precise (think
plasma displays). In fact, today's concept of "resolution" will be
largely obsolete. Get ready for pay-per-view Webcasts.
17
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
These and other obstacles have led scientists to seek answers in light itself.
Light does not have the time response limitations of electronics, does not need
insulators, and can even send dozens or hundreds of photon signal streams
simultaneously using different color frequencies. Those are immune to
electromagnetic interference, and free from electrical short circuits. They have
low-loss transmission and provide large bandwidth; i.e. multiplexing capability,
capable of communicating several channels in parallel without interference. They
are capable of propagating signals within the same or adjacent fibers with
19
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Optical computing was a hot research area in the 1980s. But the work
tapered off because of materials limitations that seemed to prevent optochips from
getting small enough and cheap enough to be more than laboratory curiosities.
Now, optical computers are back with advances in self-assembled conducting
organic polymers that promise super-tiny all-optical chips.
20
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
as well as significant shrinkage in their size and cost. An optical desktop computer
could be capable of processing data up to 100,000 times faster than current models
because multiple operations can be performed simultaneously. Other advantages of
optics include low manufacturing costs, immunity to electromagnetic interference,
a tolerance for lowloss transmissions, freedom from short electrical circuits and the
capability to supply large bandwidth and propagate signals within the same or
adjacent fibers without interference.
One oversimplified example may help to appreciate
the difference between optical and electronic parallelism. Consider an imaging
system with 1000 t 1000 independent points per mm2 in the object plane which
are connected optically by a lens to a corresponding number of points per mm2 in
the image plane; the lens effectively performs an FFT of the image plane in real
time. For this to be accomplished electrically, a million operations are required.
Parallelism, when associated with fast switching speeds, would result in staggering
computational speeds. Assume, for example, there are only 100 million gates on a
chip, much less than what was mentioned earlier (optical integration is still in its
infancy compared to electronics). Further, conservatively assume that Optical
technology promises massive upgrades in the efficiency and speed of computers,
as well as significant shrinkage in their size and cost.
An optical desktop computer could be capable of
processing data up to 100,000 times faster than current models because multiple
operations can be performedsimultaneously. Each gate operates with a switching
time of only 1 nanosecond(organic optical switches can switch at sub-picosecond
rates compared to maximum picosecond switching times for electronic switching).
Such a system could perform more than 1017 bit operations per second. Compare
this to the gigabits (109) or terabits (1012) per second rates which electronics are
either currently limited to, or hoping to achieve. In other words, a computation that
might require one hundred thousand hours (more than 11 years) of a conventional
computer time could require less than one hour by an optical one. But building an
optical computer will not be easy. A major challenge is finding materials that can
be mass produced yet consume little power; for this reason, optical computers may
not hit the consumer market for 10 to 15 years.
Another of the typical problems optical computers
have faced is that the digital optical devices have practical limits of eight to eleven
bits of accuracy in basic operations due to, e.g., intensity fluctuations. Recent
research has shown ways around this difficulty. Thus, for example, digital
partitioning algorithms, that can break matrix-vector products into lower-accuracy
sub-products, working in tandem with error-correction codes, can substantially
improve the accuracy of optical computing operations. Nevertheless, many
problems in developing appropriate materials and devices must be overcome
21
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
before digital optical computers will be in widespread commercial use. In the near
term, at least, optical computers will most likely be hybrid optical/electronic
systems that use electronic circuits to preprocess input data for computation and to
post-process output data for error correction before outputting the results.
The promise of all-optical computing remains highly
attractive, however, and the goal of developing optical computers continues to be a
worthy one. Nevertheless, many scientists feel that an all-optical computer will not
be the computer of the future; instead optoelectronic computers will rule where the
advantages of both electronics and optics will be used. Optical computing can also
be linked intrinsically to quantum computing. Each photon is a quantum of a wave
function describing the whole function. It is now possible to control atoms by
trapping single photons in small, superconducting cavities
22
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
23
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Fig.- 1
24
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
25
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Fig.- 2 (a)
26
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Fig.- 2 (b)
27
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
emitting diode (LED) under the control of a field-effect transistor (FET), can now
be made entirely out of organic materials on the same substrate for the first time. In
general, the benefit of organic over conventional semiconductor electronics is that
they should (when mass-production techniques take over) lead to cheaper, lighter,
circuitry that can be printed rather than etched. Scientists at Bell Labs have made
300-micron-wide pixels using polymer FETs and LEDs made from a sandwich of
organic materials, one of which allows electrons to flow, another which acts as
highway for holes (the absence of electrons); light is produced when electrons and
holes meet. The pixels are quite potent, with a brightness of about 2300
candela/m2, compared to a figure of 100 for present flat-panel displays . A
Cambridge University group has also made an all-organic device, not as bright as
the Bell Labs version, but easier to make on a large scale .
28
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
8
ROLE OF NLO IN OPTICAL COMPUTING
enable better and more efficient optical materials. Although organic materials have
many features that make them desirable for use in optical devices, such as high
nonlinearities, Flexibility of molecular design, and damage resistance to optical
radiation, their use in devices has been hindered by processing difficulties for
crystals and thin films. Our focus is on a couple of these materials, which have
undergone some investigation in the NASA/MSFC laboratories, and were also
processed in space either by the MSFC group, or others. These materials belong to
the classes of phthalocyanines and polydiacetylenes. These classes of organic
compounds are promising for optical thin films and waveguides. Phthalocyanines
are large ring-structured porophyrins for which large and ultrafast nonlinearities
have been observed. These compounds exhibit strong electronic transitions in the
visible region and have high chemical and thermal stability up to 400°C. We
measured the third order susceptibility of phthalocyanine, which is a measure of its
nonlinear efficiency to be more than a million times larger than that of the standard
material, carbon disulfide. This class of materials has good potential for
commercial device applications, and has been used as a photosensitive organic
material, and for photovoltiac, photoconductive, and photoelectrochemical
applications.
fiber filled with polydiacetylene. Nd:YAG green picosecond laser pulse was sent
collinearly with red cw He-Ne laser onto one end of the fiber. At the other end of
the fiber a lens was focusing the output on to the narrow slit of a monochrometer
with its grating set for the red He-Ne laser. When both He-Ne laser and Nd:YAG
laser are present there will be no output at the oscilloscope. If either one or none
of the laser beams are present we get the output at the oscilloscope showing
NAND function.
11
OPTICAL MEMORY
In optical computing two types of memory are discussed. One consists of
arrays of one-bit-store elements and other is mass storage, which is implemented
by optical disks or by holographic storage systems. This type of memory promises
very high capacity and storage density. The primary benefits offered by
holographic optical data storage over current storage technologies include
significantly higher storage capacities and faster read-out rates. This research is
expected to lead to compact, high capacity, rapid-and random-access, and low
power and low cost data storage devices necessary for future intelligent spacecraft.
The SLMs are used in optical data storage applications. These devices are used to
write data into the optical storage medium at high speed. More conventional
approaches to holographic storage use ion doped lithium niobate crystals to store
pages of data.
For audio recordings ,a 150MBminidisk with a 2.5- in diameter has been
developed that uses special compression to shrink a standard CD’s640-MB storage
capacity onto the smaller polymer substrate. It is rewritable and uses magnetic field
modulation on optical material. The mini disc uses one of the two methods to write
information on to an optical disk. With the mini disk a magnetic field placed
behind the optical disk is modulated while the intensity of the writing laser is held
constant. By switching the polarity of the magnetic field while the laser creates a
state of flux in the optical material digital data can be recorded on a single layer.
As with all optical storage media a read laser retrieves the data.
Fiber Optics: -
32
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Definition:
A basic fiber optic system consists of a transmitting device, which generates
the light signal; an optical fiber cable, which carries the light; and a receiver, which
accepts the light signal transmitted. The fiber itself is passive and does not contain
any active, generative properties.
History:
Many individuals throughout the history of the world have recognized the
value of using light to to communicate. Early defense warning systems were set up
on the Great wall of China with signal fires to warn of enemies approaching. In the
late 1700's the "optical telegraph" was invented by a French engineer named
Claude Chappe which, similar to the fire signals, used semaphores mounted on
towers, where human operators relayed messages from one tower to the next. In
1870, John Tyndal demonstrated the principle of total internal reflection by shining
a light into a water tank, poking a hole in the side, and as the water ran out in an
arc, the light took the shape and followed the water down. Ten years later,
Alexander Graham Bell patented an optical telephone system "Photophone" which
he imagined sound waves carried by light. It wasn't until many years later through
numerous advances in thinking and technical discovery's that Tyndal's and Bell's
basic concepts came together to what we now know as fiber optics. Through the
invention of the continuouswave helium-neon laser and enhancements to optical
fiber, researchers Dr. Robert Maurer, Peter Schultz, and Donald Keck of Corning
Incorporated lead the way in development of Silica manufactured fiber optics and
in 1970 were successful in manufacturing 20dB/km, cable that was tested and used
successfully in Britain. Today optical fiber is manufactured at .25dB/km, which is
an indicator of the purity of the silica and how much loss of light occurs over
distance.
Technical Info:
33
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Fiber optic systems can carry both analog and digital signals over
light waves. A system consists of a signal generator, (e.g. computer, video, audio)
an encoder, a fiber optic cable, and a decoder, and a receiving device (e.g. tv,
computer network, etc.) Fiber optics have many advantages over copper cable.
They have become a desired standard for networking backbones and hubs because
of the advantages they have over copper to achieve the speed and bandwidth
capacity. A single fiber optic cable can transmit the same amount of data as
approximately 600 pair traditional copper telecommunications wire, an transmit
data further with less boosting of the signal, it is not effected by electrical
anomalies such as lightning, it is small, light weight and easy to install.
Year2000:
34
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
With the highly purified and streamlined manufacturing process, the current
speeds of data transfer are around 5millionbps. The biggest challenge remaining is
the economic challenge. Today telephone and cable television companies generally
bring in fiber links (backbones)to remote sites serving many customers, but then
use twisted wire pair or coaxial cables from optical network units to individual
homes. This technology is often referred to "broadband" and is becoming
increasingly popular, but considerably limited to the potential of complete fiber
optic networks directly linked to individual homes. Only time will tell how long it
will take before the technology becomes reasonably economical and enough
demand is given to take that next step.
1
Uses of Optics in Computing
35
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
through the loop changes. Further, high frequency or fast switching pulses will
cause interference in neighboring wires.
On the other hand, signals in adjacent optical fibers or in optical integrated
channels do not affect one another nor do they pick up noise due to loops. Finally,
optical materials possess superior storage density and accessibility over magnetic
materials. The field of optical computing is progressing rapidly and shows many
dramatic opportunities for overcoming the limitations described earlier for current
electronic computers. The process is already underway whereby optical devices
have been incorporated into many computing systems. Laser diodes as sources of
coherent light have dropped rapidly in price due to mass production.
Also, optical CD-ROM discs are now very common in home and office
computers. Current trends in optical computing emphasize communications, for
example the use of free-space optical interconnects as a potential solution to
alleviate bottlenecks experienced in electronic architectures, including loss of
communication efficiency in multiprocessors and difficulty of scaling down the IC
technology to sub-micron levels. Light beams can travel very close to each other,
and even intersect, without observable or measurable generation of unwanted
signals. Therefore, dense arrays of interconnects can be built using optical systems.
In addition, risk of noise is further reduced, as light is immune to electromagnetic
interferences. Finally, as light travels fast and it has extremely large spatial
bandwidth and physical channel density, it appears to be an excellent media for
information transport and hence can be harnessed for data processing. This high
bandwidth capability offers a great deal of architectural advantage and flexibility.
Based on the technology now available, future systems could have 1024 smart
pixels per chip with each channel clocked at 200MHz (a chip I/O of 200Gbits per
second), giving aggregate data capacity in the parallel optical highway of more
that 200Tbits per second; this could be further increased to 1000Tbits. Free-space
optical techniques are also used in scalable crossbar systems, which allow arbitrary
interconnections between a set of inputs and a set of outputs. Optical sorting and
optical crossbar inter-connects are used in asynchronous transfer modes or packet
routing and in shared memory multiprocessor systems.
In optical computing two types of memory are discussed. One consists of
arrays of one-bit-store elements and the other is mass storage, which is
implemented by optical disks or by holographic storage systems. This type of
memory promises very high capacity and storage density. The primary benefits
offered by holographic optical data storage over current storage technologies
include significantly higher storage capacities and faster read-out rates. This
research is expected to lead to compact, high-capacity, rapid- and random-access,
radiation-resistant, low-power, and low-cost data storage devices necessary for
future intelligent spacecraft, as well as to massive-capacity and fast-access
36
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
terrestrial data archives. As multimedia applications and services become more and
more prevalent, entertainment and data storage companies are looking at ways to
increase the amount of stored data and reduce the time it takes to get that data out
of storage. The SLMs and the linear array beam steerer are used in optical data
storage applications. These devices are used to write data into the optical storage
medium at high speed.
The analog nature of these devices means that data can be stored at much
higher density than data written by conventional devices. Researchers around the
world are evaluating a number of inventive ways to store optical data while
improving the performance and capacity of existing optical disk technology. While
these approaches vary in materials and methods, they do share a common
objective: expanded capacity through stacking layers of optical material. For audio
recordings, a 150-MB minidisk with a 2.5-in. diameter has been developed that
uses special compression to shrink a standard CD’s 640-MB storage capacity
onto the smaller polymer substrate. It is rewritable and uses magnetic field
modulation on optical material. The minidisk uses one of two methods to write
information onto an optical disk. With the minidisk, a magnetic field placed behind
the optical disk is modulated while the intensity of the writing laser head is held
constant. By switching the polarity of the magnetic field while the laser creates a
state of flux in the optical material, digital data can be recorded on a single layer.
As with all optical storage media, a read laser retrieves the data. Along with
minidisk developments, standard magneto-optical CD technology has expanded the
capacity of the 3.5-in. diameter disk from 640 MB to commercially available 1 GB
storage media. These conventional storage media modulate the laser instead of the
magnetic field during the writing process. Fourth-generation 8,5.25 in.diameter
disks that use the same technology have reached capacities of 4 GB per disk. These
disks are used mainly in ‘jukebox’ devices. Not to be confused with the musical
jukebox, these machines contain multiple disks for storage and backup of large
amounts of data that need to be accessed quickly.
Beyond these existing systems are several laboratory systems that use
multiple layers of optical material on a single disk. The one with the largest
capacity, magnetic super-resolution (MSR), uses two layers of optical material.
The data is written onto the bottom layer through a writing laser and magnetic field
modulation (MFM). When reading the disk in MSR mode, the data is copied from
the lower layer to the upper layer with greater spacing between bits. In this way,
data can be stored much closer together (at distances smaller than the read beam
wavelength) on the bottom layer without losing data due to averaging across bits.
This method is close to commercial production, offering capacities of up to 20 GB
on a 5.25 in. disk without the need for altering conventional read-laser technology.
37
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Advanced storage magnetic optics (ASMO) builds on MSR, but with one
exception.
Standard optical disks, including those used in MSR, have grooves and lands
just like a phonograph record. These grooves are used as guideposts for the writing
and reading lasers. However, standard systems only record data in the grooves, not
on the lands, wasting a certain amount of the optical material’s capacity. ASMO
records data on both lands and grooves and, by choosing groove depths
approximately 1/6 the wavelength of the reading laser light, the system can
eliminate the crosstrack crosstalk that would normally be the result of recording on
both grooves and lands. Even conventional CD recordings pick up data from
neighboring tracks, but this information is filtered out, reducing the signal-to-noise
ratio. By closely controlling the groove depth, ASMO eliminates this problem
while maximizing the signal-to-noise ratio. MSR and ASMO technologies are
expected to produce removable optical disk drives with capacities between 6 and
20 GB on a 12-cm optical disk, which is the same size as a standard CD that holds
640 MB. Magnetic amplifying magneto-optical systems (MAMMOS) use a
standard polymer disk with two or three magnetic layers. In general terms,
MAMMOS is similar to MSR, except that when the data is copied from the bottom
to the upper layer, it is expanded in size, amplifying the signal. According to
Archie Smith of Storagetek’s Advanced Technology Office (Louisville, CO),
MAMMOS represents a two-fold increase in storage capacity over ASMO.
Technology developed by Call/Recall Inc. (San Diego, CA) could help bridge the
gap between optical disk drives and holographic memories. Called 2-photon
optical storage technology (which got its start with the assistance of the Air Force
research laboratories and DARPA), the Call/Recall systems under development use
a single beam to write the data in either optical disks with up to 120 layers, or into
100-layer cubes of active-molecule-doped MMA polymer. In operation, a mask
representing data is illuminated by a mode-locked Nd:YAG laser emitting at 1064
nm with pulse durations of 35 ps. The focal point of the beam intersects a second
beam formed by the second harmonic of the same beam at 532 nm. The second
beam fixes the data spatially and temporally. A third beam from a He Ne laser
emitting at 543 nm reads the data by causing the material to fluoresce. The
fluorescence is read by a chargecoupled device (CCD) chip and converted through
proprietary algorithms back into data. Newer versions of the system use a
Ti:Sapphire laser with 200-fs pulses. Call/Recall’s Fredrick McCormick said the
newer and older approaches offer different strengths. The YAG system can deliver
higher-power pulses capable of storing megabits of data with a single pulse, but at
much lower repetition rates than the Ti:Sapphire laser with its lower-power pulses.
Thus, it is a trade-off. Call/Recall has demonstrated the system using portable
apparatus comprised of a simple stepper-motor-driven stage and 200-microwatt
38
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
HeNe laser in conjunction with a low-cost video camera. The company estimates
that an optimized system could produce static bit error rates (BER) of less than 9
10–13. McCormick believes that a final prototype operating at standard CD
rotation rates would offer BERs that match or slightly exceed conventional optical
disk technology. Researchers such as Demetri Psaltis and associates at the
California Institute of Technology are also using active-molecule-doped polymers
to store optical data holographically.
Their system uses a thin polymer layer of PMMA doped with
phenanthrenequinone (PQ). When illuminated with two coherent beams, the
subsequent interference pattern causes the PQ molecules to bond to the PMMA
host matrix to a greater extent in brighter areas and to a lesser extent in areas where
the intensity drops due to destructive interference. As a result, a pair of partially
offsetting index gratings is formed in the PMMA matrix. After writing the
hologram into the polymer material, the substrate is baked, which causes the
remaining unbounded PQ molecules to diffuse throughout the polymer, removing
the offsetting grating and leaving the hologram. A uniform illumination is the final
step, bonding the diffuse PQ throughout the matrix and fixing the hologram in the
polymer material.
Storagetek’s Archie Smith estimates that devices based on this method could
hold between 100 and 200 GB of data on a 5.25-in diameter polymer disk.
More conventional approaches to holographic storage use irondoped
lithium niobate crystals to store pages of data. Unlike standard magneto-optical
storage devices, however, the systems developed by Pericles Mitkas at Colorado
State University use the associative search capabilities of holographic memories.
Associative or content-based data access enables the search of the entire memory
space in parallel for the presence of a keyword or search argument. Conventional
systems use memory addresses to track data and retrieve the data at that location
when requested. Several applications can benefit from this mode of operation
including management of large multimedia databases, video indexing, image
recognition, and data mining.
Different types of data such as formatted and unformatted text, gray scale
and binary images, video frames, alphanumeric data tables, and time signals can be
interleaved in the same medium and we can search the memory with either data
type. The system uses a data and a reference beam to create a hologram on one
plane inside the lithium niobate. By changing the angle of the reference beam,
more data can be written into the cube just like pages in a book. The current
systems have stored up to 1000 pages per spatial location in either VGA or VGA
resolutions. To search the data, a binary or analog pattern that represents the search
argument is loaded into a spatial light modulator and modulates a laser beam. The
light diffracted by the holographic cube on a CCD generates a signal that indicates
39
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
the pages that match the sought data. Recent results have shown the system can
find the correct data 75 percent of the time when using patterns as small as 1 to 5
percent of the total page. That level goes up to 95 to 100 percent by increasing the
amount of data included in the search argument.2
40
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
lithium niobate crystal-based devices. The California team has been working to
incorporate their material into a working prototype. Development of such a device
could revolutionize the information superhighway and speed data processing for
optical computing. Another group at Brown University and the IBM.
Almaden Research Center (San Jose, CA) have used ultrafast laser pulses to
build ultrafast datastorage devices. This group was able to achieve ultrafast
switching down to 100ps. Their results are almost ten times faster than currently
available Òspeed limitsÓ. Optoelectronic technologies for optical computers and
communication hold promise for transmitting data as short as the space
between computer chips or as long as the orbital distance between satellites. A
European collaborative effort demonstrated a high-speed optical data input and
output in free-space between IC chips in computers at a rate of more than 1 Tb/s.
Astro Terra, in collaboration with Jet Propulsion Laboratory (Pasadena, CA) has
built a 32-channel 1-Ggb/s earth Ðto Ðsatellite link with a 2000 km range. Many
more active devices in development, and some are likely to become crucial
components in future optical computer and networks.
The race is on with foreign competitors. NEC (Tokyo, Japan) have
developed a method for interconnecting circuit boards optically using Vertical
Cavity Surface Emitting Laser arrays (VCSEL). Researchers at Osaka City
University (Osaka, Japan) reported on a method for automatic alignment of a set of
optical beams in space with a set of optical fibers.
As of last year, researchers at NTT (Tokyo, Japan) have designed an optical
back plane with free Ðspace optical interconnects using tunable beam deflectors
and a mirror. The project had achieved 1000 interconnections per printed-circuit
board, with throughput ranging from 1 to 10 Tb/s.
Optics has a higher bandwidth capacity over electronics, which enables more
information to be carried and data to be processed arises because electronic
communication along wires requires charging of a capacitor that depends on
length. In contrast, optical signals in optical fibers, optical integrated circuits, and
free space do not have to charge a capacitor and are therefore faster.
Another advantage of optical methods over electronic ones for computing is
that optical data processing can be done much easier and less expensive in parallel
than can be done in electronics. Parallelism is the capability of the system to
execute more than one operation simultaneously. Electronic computer architecture
is, in general, sequential, where the instructions are implemented in sequence. This
implies that parallelism with electronics is difficult to construct. Parallelism first
appeared in Cray super computers in the early 1980s.
Two processors were used in conjunction with the computer memory to
achieve parallelism and to enhance the speed to more than 10 Gb/ s. It was later
realized that more processors were not necessary to increase computational speed,
41
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
but could be in fact detrimental. This is because as more processors are used, there
is more time lost in communication. On the other hand, using a simple optical
design, an array of pixels can be transferred simultaneously in parallel from one
point to another. To appreciate the difference between both optical parallelism and
electronic one can think of an imaging system of as many as 1000x1000
independent points per mmin the object plane which are connected optically by a
lens to a corresponding 1000x 1000 points per mm in the image plane. For this to
be accomplished electrically, a million nonintersecting and properly isolated
conduction channels per mm would be required.
Parallelism, therefore, when associated with fast switching speeds, would
result in staggering computational speeds. Assume, for example, there are only 100
million gates on a chip, much less than what was mentioned earlier (optical
integration is still in its infancy compared to electronics). Further, conservatively
assume that each gate operates with a switching time of only 1 nanosecond
(organic optical switches can switch at sub-picosecond rates compared to
maximum picosecond switching times for electronic switching). Such a system
could perform more than 1017 bit operations per second. Compare this to the
gigabits (109) or terabits (1012) per 6 second rates which electronics are either
currently limited to, or hoping to achieve.
In other words, a computation that might require one hundred thousand
hours (more than 11 years) of a conventional computer could require less than one
hour by an optical one.
Another advantage of light results because photons are uncharged and do not
interact with one another as readily as electrons. Consequently, light beams may
pass through one another in fullduplex operation, for example without distorting
the information carried. In the case of electronics, loops usually generate noise
voltage spikes whenever the electromagnetic fields through the loop changes.
Further, high frequency or fast switching pulses will cause interference in
neighboring wires. Signals in adjacent fibers or in optical integrated channels do
not affect one another nor do they pick up noise due to loops. Finally, optical
materials possess superior storage density and accessibility over magnetic
materials.
Obviously, the field of optical computing is progressing rapidly and shows
many dramaticopportunities for overcoming the limitations described earlier for
current electronic computers.
The process is already underway whereby optical devices have been
incorporated into many computing systems. Laser diodes as sources of coherent
light have dropped rapidly in price due to mass production. Also, optical CD-ROM
discs have been very common in home and office computers.
42
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
OPTICAL DISK
13
WORKING
The 780nm light emitted from AlGaAs/GaAs laser diodes is collimated by a lens
and focused to a diameter of about 1micrometer on the disk. If there is no pit where
the light is incident, it is reflected at the Al mirror of the disk and returns to the
lens, the depth of the pit is set at a value such that the difference between the path
of the light reflected at a pit and the path of light reflected at a mirror is an integral
multiple of half-wavelength consequently, if there is a pit where light is incident,
the amount of reflected light decreases tremendously because the reflected lights
are almost cancelled by interference. The incident and reflected beams pass
through the quarter wave plate and all reflected light is introduced to the
photodiode by the beam splitter because of the polarization rotation due to the
quarter wave plate. By the photodiode the reflected light, which has a signal
whether, a pit is on the disk or not is changed into an electrical signal.
43
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
As Moore's law keeps giving us faster and faster computers, chip builders
also need higher-bandwidth data connections. But excessive heat dissipation and
power requirements make conventional wires impractical at higher frequencies,
which has lead researchers to develop new ways to store, transmit and elaborate
optically-encoded information.
The solution found by the MIT team and detailed in a paper published in the
journal Optics Letters is notable not only because it achieves these objectives, but
also because it changes the way physicists have been looking at a class of materials
that were previously thought to be unsuitable for manufacturing lasers.
The second strategy was to "strain" the germanium, pulling its atoms slightly
farther apart than they would be naturally by growing it directly on top of a layer
of silicon. This makes it easier for electrons to jump into the photon-emitting state.
lasers, making them more attractive as sources of light for optical data
connections and, one day, for computing as well.
One of the issues of current chip design is the excessive power needed to
transport and store ever increasing amounts of data. A possible solution is to use
optics not just for sending data, but also to store information and perform
calculations, which would reduce heat dissipation and increase operating speeds.
Disproving previous beliefs in the matter, MIT researchers have demonstrated the
first laser built from germanium which can perform optical communications... and
it's also cheap to manufacture.
As Moore's law keeps giving us faster and faster computers, chip builders
also need higher-bandwidth data connections. But excessive heat dissipation and
power requirements make conventional wires impractical at higher frequencies,
45
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
which has lead researchers to develop new ways to store, transmit and elaborate
optically-encoded information.
The solution found by the MIT team and detailed in a paper published in the
journal Optics Letters is notable not only because it achieves these objectives, but
also because it changes the way physicists have been looking at a class of materials
that were previously thought to be unsuitable for manufacturing lasers.
The second strategy was to "strain" the germanium, pulling its atoms slightly
farther apart than they would be naturally by growing it directly on top of a layer
of silicon. This makes it easier for electrons to jump into the photon-emitting state.
46
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
NASA scientists are working to solve the need for computer speed using
light itself to accelerate calculations and increase data bandwidth.
47
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
(billionths) for logic operations - and picoseconds (trillionths!) for the switches and
gates in chips.
"That's great in theory," says Dr. Donald Frazier of NASA's Marshall Space
Flight Center. "Except that electronic signals, even with Very Large Scale
Integration (VLSI) and maximum miniaturization, are bogged down by many
aspects of the solid materials they travel through. So we've had to find a faster
medium for the signals - and the answer seems to be light itself!" Above: Dr.
Donald Frazier monitors a blue laser light used with electro-optical materials.
Light travels at 186,000 miles per second. That's 982,080,000 feet per
second -- or 11,784,960,000 inches. In a billionth of a second, one nanosecond,
photons of light travel just a bit less than a foot, not considering resistance in air or
of an optical fiber strand or thin film. Just right for doing things very quickly in
microminiaturized computer chips.
"Entirely optical computers are still some time in the future," says Dr.
Frazier, "but electro-optical hybrids have been possible since 1978, when it was
learned that photons can respond to electrons through media such as lithium
niobate. Newer advances have produced a variety of thin films and optical fibers
that make optical interconnections and devices practical. We are focusing on thin
films made of organic molecules, which are more light sensitive than inorganics.
Organics can perform functions such as switching, signal processing and frequency
doubling using less power than inorganics. Inorganics such as silicon used with
organic materials let us use both photons and electrons in current hybrid systems,
which will eventually lead to all-optical computer systems."
Logic gates are the building blocks of any digital system," he continues.
"An optical logic gate is a switch that controls one light beam with another. It is
"on" when the device transmits light, and "off" when it blocks the light."
synchronously with the pump beam, showing the characteristic table of an AND
logic gate."
"Our setup for the picosecond switch was similar, except that the
phthalocyanine film was replaced with a hollow fiber coated from inside with a
thin polydiacetylene film. Both collinear laser beams were focused on one end of
the tube, and a lens at the other end focused the output onto a monochrometer with
a fast detector attached. The product of the two beams demonstrates three of the
four characteristics of a NAND logic gate."
"Optical bistable devices and logic gates such as these are the equivalent of
electronic transistors," concludes Dr. Abdeldayem. "They operate as very high
speed on-off switches and are also useful as optical cells for information storage."
Logic gates are the building blocks of any digital system. An optical logic
gate is a switch that controls one light beam by another; it is ÒONÓ when the
device transmits light and it is ÒOFFÓ when it blocks the light. Recently we
demonstrated in our laboratory at NASA/Marshall Space Flight Center two fast all-
optical switches using phthalocyanine thin films and polydiacetylene fiber. The
phthalocyanine switch is in the nanosecond regime and functions as an all-optical
AND logic gate, while the polydiacetylene one is in the picosecond regime and
exhibits a partial all-optical NAND logic gate.
To demonstrate the AND gate in the phthalocyanine film, we waveguided
two focused collinear beams through a thin film of metal-free phthalocyanine film.
The film thickness was ~ 1 m and a few millimeters in length. We used the second
50
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
(b) dynamically assigning at least one bus time slot to given ones of a
plurality of computer sub-system elements during a single computer system clock
signal cycle;
2. The method of claim 1, wherein step (c) includes: applying a signal from
one of said plurality of computer sub-system elements to said computer bus during
said at least one bus time slot, a plurality of dynamically assigned bus time slots
being divisible among said plurality of computer sub-system elements in
accordance with a bandwidth requirement of one of the computer sub-system
elements.
4. The method of claim 1, wherein step (c) includes a step of converting said
signal from said computer sub-system element to a light signal.
5. The method of claim 4, wherein step (c) includes coupling said light
signal to an optical computer bus.
53
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Processor-memory buses are often design-specific, wile both I/O buses and
backplane buses are frequently standard buses with parameters established by
industry standards. The distinction between bus types is becoming increasingly
difficult to specify. Thus, the present application generically refers to computer
buses to encompass all processor-memory buses, I/O buses, and backplane buses.
The physical operation and constraints of existing computer bus designs are
most fully appreciated .A computer bus 20 positioned on a backplane 22. The
computer bus 20 is a set of wires, effectively forming a transmission line. A
random number of system cards (or cards) 24A-24N are attached to the computer
54
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
bus 20. By way of example, the cards 24 may include a video processing card, a
memory controller card, an I/O controller card, and a network card. Each card 24 is
connected to the computer bus 20 through a connector 26. Thus, each card 24 is
electrically connected to the set of wires forming the computer bus 20. As a result,
one card, say card 24A, can communicate with another card, say card 24N, by
writing information onto the computer bus 20. Only one card 24 can write
information onto the computer bus 20 at a time, thus a computer bus 20 can
generate a performance bottleneck as different cards 24 wait to write information
onto the bus 20.
Another problem associated with a traditional computer bus 20, is that its
performance is constrained by complicated electrical phenomenon. For example,
the connectors 26 effectively divide the bus into transmission line segments,
resulting in complicated transmission line effects. Note that the transmission line
segments will vary depending upon the number of cards 26 connected to the bus
20. This periodic loading of the bus 20 makes it difficult to optimize bus
performance. In addition, each connector 26 produces a lumped discontinuity with
parallel capacitance and series inductance, thereby complicating the electrical
characteristics of the bus 20. Note also that "T-connections" are formed between
the wires of a computer bus 20 and the wires to a connector 26. The T-connections
complicate the electrical characteristics of the computer bus 20.
The optical computer bus is extremely fast, with bus signals moving at
approximately the speed of light ((index of refraction of the fiber)31 1×the speed
of light). The operation of the bus is not compromised by transmission line effects
associated with prior art computer buses. Further, the optical computer bus of the
invention does not suffer from electrical noise problems. The optical computer bus
is compact and is therefore ideal for space-constrained modern computers. Despite
its radically different design and configuration, the computer bus of the invention
otherwise operates in a standard manner. Thus, the computer bus can be used in
56
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
existing computers and system designers can still rely upon known bus design
techniques.
A digital gate computer bus 40, also called a chip bus, in accordance with
the invention. The chip bus 40 of the invention uses digital circuits 30 to perform
the function executed by a conventional computer bus. That is, the chip bus 40 of
the invention is used to perform a set of logical OR operations with digital gates so
that these operation do not have to be performed as wired OR operations on the
wires of a computer bus. In this way, the transmission line problems associated
with prior art computer buses are eliminated.
The operation of the invention is more fully appreciated with a simple example.
Typically, each card attached to a computer bus has N communication bits
corresponding to the N wires forming the computer bus. Thus, for example, if four
cards are attached to a computer bus, then each card has a designated bit that reads
and writes signals to a designated wire of the computer bus. If any card on the bus
writes a digital ONE to this designated wire of the computer bus, then all cards on
the bus read a digital high signal for this designated bit. This is a logical OR
operation performed by a hardwired circuit (the wire of the bus). The present
invention eliminates the physical wires of traditional computer buses and executes
the operation associated with such wires with digital gates. That is, the chip bus 40
of the invention performs logical OR operations with digital gates in order to
eliminate the transmission line problems associated with prior art computer buses.
57
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
includes a card signal driver 52 with a bus input signal driver 54, implemented as
an inverter, and a bus output signal driver 56, also implemented as an inverter.
Thus, it can be appreciated that the chip bus bit processor 50 of receives a
single bit input signal from four cards (49A, 49B, 49C, 49D). In particular, each
single bit input signal is driven by the bus input signal driver 54 and applied to the
logical OR circuit 51. If any single bit input signal is a digital ONE on the logical
OR circuit 51, then a high output is generated at all output nodes. For the
embodiment, the high output signal is seen by the card logic circuit 66 after
processing by inverters 56 and 64.
In one embodiment of the invention, the card 49B may include a card
transceiver 60B. In this embodiment, the card transceiver 60B includes a logic
output signal driver 62, implemented as an inverter, and a logic input signal driver
64, also implemented as an inverter. The signals from the card transceiver 60B are
then processed by a logic circuit 66 in a conventional manner.
Therein are the same components shown , but the components are
rearranged to more fully describe the invention. In addition, the logical OR circuit
51 as being implemented with a four input OR gate. Thus, it is seen that each card
(49A, 49B, 49C, 49D) generates a single bit signal that is respectively applied to
the chip bus input lines (44A, 44B, 44C, 44D). The four signals are routed to the
four input OR gate 51. The output of the four input OR gate 51 is then routed back
to the cards (49A, 49B, 49C, 49D) through their respective chip bus output lines
(46A, 46B, 46C, 46D).
A four bit digital gate computer bus in accordance with the invention. The
four bit digital gate computer bus is used in conjunction with four processing cards
(49A, 49B, 49C, 49D). The four bit digital gate computer bus includes a chip bus
package 70 with package pins 72. Standard packaging techniques may be used to
form this structure. Within the package 70 are four chip bus bit processors (50A,
50B, 50C, 50D). The package 70 is positioned on a backplane 22.
Each processing card (49A, 49B, 49C, 49D) generates a single bit signal that
is applied to one of the chip bus bit processors 50. In particular, each processing
card generates a single bit signal that is applied to a chip bus input line 44 formed
on backplane 22. The signal reaches a package pin 72 and is then routed to a chip
bus bit processor 50 via a package internal trace 74. After processing by the chip
bus bit processor 50 is completed, the output signals are applied to chip bus output
58
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
lines 46 formed in the backplane 22. The chip bus output lines 46 route the output
signals to their respective cards for processing in a standard fashion.
The invention has now been fully described. Attention presently turns to a
discussion of various implementation issues. Implementations of the chip bus 40 of
the invention will have the shared portion of a physical bus implemented with
digital gates and will use point-to-point wiring to connect the daughterboards
(cards 24). As used herein, point-to-point wiring refers to wiring running directly
between pins of two packages, without "T-connections", "Y-connections", or
related configurations or sources which complicate signal transmission.
The preferred embodiment of the invention uses separate chip bus input lines
44 and chip bus output lines 46. However, it is possible to use bidirectional wires
to make these connections. The bidirectional wires save a factor of two in signals,
but cannot reach the speed attainable by the unidirectional technique, unless
special transceivers are used that can simultaneously send and receive on the same
line. For the highest speed systems, it may be advantageous to use differential
simultaneous bidirectional signaling to reduce system noise.
If the electrical distance from the card 49 to the chip bus 40 is less than
about half the transition time, the signal can be unterminated, and the driver can be
quite small. If the line is long enough to be terminated, it is possible to operate in
59
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
the 50 to 100 Ohm regime, rather than the sub-20 Ohm regime associated with a
heavily loaded conventional bus. Note that the termination can be done by
correctly sizing the driver transistors.
One way to use the chip bus 40 is as a drop-in replacement for a traditional
bus structure. In this mode, the chip bus provides the advantages of lower power
because it is easier to drive the lines, there are smaller propagation delays because
the point-to-point wiring is not periodically loaded, and the bus topology is
decoupled from the electrical behavior. The chip bus does not suffer from the
multiple reflection noise and settling delays associated with classical bus
implementations.
The chip bus 40 of the invention is extremely fast. Simulated chip bus 40
designs have shown bit rates of 2.4 Gbits/sec per line. The delay through the chip
bus 40 is only 330 pS. A portion of the chip bus's speed is attributable to the fact
that input signals to the bus 40 can be pipelined, four input signals A, B, C, and D
are respectively carried by chip bus input lines 44A, 44B, 44C, and 44D at time
To. The pulse width of each signal is equivalent to the pulse width of the signal
clock, shown as Tp.. 6B illustrates the progression of the four input signals after a
clock cycle, that is, at time T=To Tp. the same signals on the chip bus output lines
46A, 46B, 46C, and 46D. The signals appear on the chip bus output lines at a time
T=To nTp, where n is the number of clock cycles required to drive the signals
through the chip bus 40. that it is possible to have an input signal to the bus and an
output signal from the bus every cycle. This pipelining capability results in
extremely high processing speeds that are not possible with traditional bus
architectures.
Current processor designs have about twenty gates between latches. The sum
of the setup and hold time of the latches is around 10% of the cycle time, or two
unit gate delays. The bus chip can be modeled as a pure delay, it doesn't change
pulse width. This implies that up to ten bus signals could be stacked in one
processor cycle. If some margin is allowed for timing tolerances, a practical limit
near eight transactions per cycle might be obtained with very careful design.
60
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
The real limitations on the speed of the system are clock skew and bit-to-bit
skew within a single package. Careful design of the wiring on the backplane 22
allows wire skew to be reduced to below all other skews in the system. Clock skew
can be kept low by using self-compensating clock drivers.
If the bus is wide enough that more than one package 70 is required, two
elements will contribute to the bit-to-bit skew. One is the difference in average
total delay between the parts, and the other is the spread in delay between the pins
within a single part. The traditional way of coping with part-to-part variations is to
bin the parts. Note that this does not cause yield loss, it just requires that any
particular board be populated with parts that have the same total delay dash-
number. It is also possible to build active compensation circuits into the parts to
force the average delay to match, for example, a reference delay printed on the
board.
The clock protocol also influences the effect of delay variations. If the
signalling is source synchronized on a chip-by-chip basis, that is each group of bits
that is carried by a single bus chip carries its own clock, the sensitivity to interchip
delay variations may be minimized. This does add some complexity to the receiver
design to ensure that all the bit groups are correctly realigned. The source clock
may be used to provide the reference input for delay lock loops to compensate
these errors. Errors in arrival time of signals at the inputs of a single bus chip can
directly subtract from the signal pulse width.
Each signal can carry its own clock, for example, by using Manchester
coding as the synchronization protocol. Any method that carries the clock on the
same line as the signal will pay some overhead in bandwidth and latency. One
advantage of using a self clocking protocol is that all inputs to a chip bus can be
individually actively delay compensated by choosing one of the inputs as a
reference for all the others. This can be made to work both for the chip buses and
the system chips, and provides a global clock synchronization method as a side
effect of minimizing skew errors in the interconnect.
61
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
The ability to swap out boards in servers without powering down the system
or stopping the clock has become a requirement for new server designs. This is
rather difficult to implement using a traditional bus structure because both the
insertion and removal of the board produces electrical transients on the backplane.
The chip bus of the invention provides an elegant solution to this problem. A
disable pin for each port on a bus chip can be provided to force the corresponding
port into an idle state where the output is not driven and the input is ignored. This
isolates a board being removed or inserted from the bus. The control of these
disable signals can be derived from variable length fingers on the backplane
connectors.
When state machines or other intelligence is not used, the chip bus is
logically equivalent to passive wires on a backplane. This allows them to run at the
maximum speed that the technology will support and permits bit-slicing the bus to
accommodate real world packaging constraints.
The pin count required for a package 70 can be reduced as far as desirable
by relying upon multiple chips. A 70 bit bus supporting 16 cards can be
implemented with 8 chip buses 40 of the type that use separate input and output
lines. Each chip bus 40 could be formed in a 432 pin package with the processor
and bus running at the same clock speed. This has the advantage of requiring no
control signals to the bus chips and can provide more bandwidth if the bus were
run at a multiple of the processor clock.
The optical bus bit processor, also called a star coupler, 80 includes a set of
N input optical fibers 82A-82N carry input signals to an optical fiber link 84,
shown in this embodiment as a fiber ring. A set of N output optical fibers 86A-86N
carry output signals. Consistent with previous embodiments of the invention, if a
single input signal is digitally ON (equivalent to a light pulse), then the fiber link
84 will cause each output optical fiber 86A-86N to carry a digitally ON signal.
Thus, the apparatus 80 performs a logical OR operation, consistent with the
previous embodiments of the invention.
63
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
A system card (not shown) is attached to a connector 96. The system card
may be memory card, a local input/output card, a network input/output card,
graphics card, etc. Thus, each system card is typically in the form of a computer
sub-system. Alternately, all computer sub-systems can be contained on a single
card.
Each optical bus bit processor 80 is capable of processing four input bits and
producing four output bits. The output signals on output fibers 86A-86D are
applied to a set of output connector cards. For the sake of simplicity, a single
output optical connector card 110. The output optical connector card 110 includes
a light signal receiver array 112, which may be implemented using a set of
64
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
The optical bus bit processor 80 is positioned on a substrate 92. The optical
bus bit processor 80 may also be implemented as a fused coupler or in free space.
In the free space embodiment, an individual light source of a set of light sources at
a sending end is capable of transmitting a signal through space. The single signal
generates an output signal at a set of light receiving sources. The light sources and
light receiving sources are controlled by dynamic bandwidth allocators of the type
described below.
Those skilled in the art will appreciate that the optical computer bus 90 of
the invention is very fast. The optical and electrical implementations of the
invention allow the bus clock to operate at a multiple of the system clock. This
operation illustrates a four word bus which makes connections to four input optical
connector cards 100, although only one is shown for the sake of simplicity.
Similarly, the bus is connected to four output optical connector card 110, although
only one is shown. Relying upon this example, if the clock for the optical bus 90 is
operated at four times the speed of the system clock, then four bus time slots exist.
Each of the four connector cards can transmit an eight bit data word in a bus time
slot.
Waveform 120 illustrates the system clock signal. Waveform 122 illustrates
the bus clock signal, which is four times faster than the system clock signal.
Waveform 124 illustrates that a first input optical connector card transmits data (an
eight bit word in this example) during the first bus time slot, which corresponds to
the first bus clock signal cycle. The second input optical connector card transmits
data during the second bus clock cycle, the third input optical connector card
transmits data during the third bus clock cycle, and the fourth input optical
connector card transmits data during the fourth bus clock cycle, during one system
clock cycle and four bus clock cycles (bus time slots), each optical connector card
is allowed to transmit data on the system bus. This process may be repeated for
subsequent clock cycles.
65
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
is applied to an input node of a flip-flop 160. The flip-flop 160 is enabled if the bus
clock signal is high and the transmission mask bit is set to a digital high value. In
this case, the logical AND gate 162 generates a digital high value, or flip-flop
enable signal, to enable the flip-flop 160. Thus, it can be appreciated that the
transmission mask bit controls the output from the transmission circuit 152. The
output from the transmission circuit 152 is used as a drive signal for the laser array
104. Preferably, a deskew circuit 154 and a drive circuit 156 are used at the output
end of the transmission circuit 152.
example, a traditional backplane 22, connectors 48, and cards 49 need not be used.
The embodiments were chosen and described in order to best explain the principles
of the invention and its practical applications, to thereby enable others skilled in
the art to best utilize the invention and various embodiments with various
modifications as are suited to the particular use contemplated.
APPLICATIONS :-
68
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
MERITS :-
69
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
DRAWBACKS
70
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
Another group at brown university and the IBM , Almaden research center
has used ultrafast laser pulses to build ultra fast data storage devices . this groupe
was able to achivie ultra fast switching down to 100 picosecond .
71
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
FUTURE TRENDS
72
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
REFERENCES
6. www.sciam.com
7. www.msfc.com
73
DEPTT. OF COMPUTER SCIENCE OPTICAL COMPUTERS
74