David Phaire
By the end of this decade, measurement of computer performance will have changed dramatically. Today, a CPUs clock frequency is the primary performance indicator. Over time operating frequency of PCs will began to play a smaller role until it becomes relatively unimportant. Imagine if Supercomputers to PCs were powered by the same microprocessor operating at the same frequency. System performance will then be a function of memory and I/O to satisfy the data requirements of the CPU. The price of the computer would correlate with the bandwidth, latency and expandability of the system. To build a balanced system, memory units must be able to deliver information quickly enough to satisfy the microprocessor, or efficiency is lost. High performance systems are differentiated not by processor performance, but by how their cache, main memory and buses are used.
Until the late 60's and early 70's the main effort of software application designers, was aimed at developing applications with higher functional characteristics. But the emergence of personal computers diverted the designers focus from functional exellence to interaction flexibility and ultimately 'user friendliness'.
Virtual Reality hype is becoming a large part of
everyday life. This report explores the components of actual
virtual reality systems, such as HMD's, gloves,spaceballs and Wands.
The industry factions have reached a compromise on
the standard for future digital videodisks (DVD) devices which
are expected to replace the CD-ROM drive. The new standard is
targeted to ultimately replace the home VCR. One problem that
has hampered its progress , is that it will be some time
before real-time compression techniques will be available to
record TV programs.
This new standard combines elements of two different
proposals. A proposal from Philips and Sony , called
multimedia CD-ROM (MMCD), specifies that future DVD devices
will be able to play current CD-ROM discs.
A consortium led by Matsushita, Time Warner and Toshiba
favours a higher capacity specification called SD(super
density) thats incompatible with current CD-ROM discs.
Backward compatibility has been incorporated into the new
standard: All DVD devices including TV set-top boxes and
computer drives will be able to read regular CD-ROMs. Also
discs for the new standard will be the same size as current CD-
ROM discs. The unified specification allows for 4.7GB of space
on one side of a disc, width is enough to hold 133mins of MPEG-
2 video. But that capacity can be quadrupled to about 18GB,
layered options. But single sided single layered CDs are likely to dominate because their capacity is quite adequate
for computer applications and run time movies. More over the
cost of producing them isnt likely to be very different from
current CD-ROM prices.
Double sided and double layered technologies are not new , but
neither have been produced in mass quantities.
Other added advantages of the DVD technology is that it can
produce a data transfer rate of about 1.2 MBps , which is
about 8 to 10 time faster than an existing quad-speed CR-ROM.
The price of this technology will be about $500 a drive ,
which is quite reasonable. This will probably hit the PC
market quicker than others and be used for games or
references.
2.2 Designing chips (Photolithography)
Since the IC was developed, the number of transistors
that engineers can pack on a chip has increased at a phenomenal rate.
ICs are made by the Photolithography process
in which the patterns of metal or chemically treated silicon are
layered one atop another, on to a die of silicon.Current lithographic
processes employ a mercury light source whose 0.365 micron wavelength
creates the 0.35 micron features.Building even smaller chip features
requires using light sources with even shorter wavelength.That means
designers had to move from visible light , to ultraviolet light, and
finally to the X-Rays territory.But using the X rays for the
photolithographic introduces a new set of problems. For example the
issue of having a reliable X ray source, the X rays cannot be
focused with optical lenses and therefore the mask, which produces the
required pattern on the silicon, must be the size of the features
themselves and furthermore the materials opaque to light are not
necessarily opaque to X rays.
Today's systems use closely-coupled cache controllers, cache SRAMs (static random access memory) and cache tags to support data requirements of CPUs. But, as processor performance continues to increase cost, asynchronous SRAMs won't meet these demands. PCs and servers, workstations each use different cache architectures: PCs use an asynchronous cache, servers and workstations rely on synchronous cache and super workstations rely on pipelined caching architectures. Pentium-based PCs run at 100 MHz internal clocks and 66-MHz external bus rates. Primary cache is on chip, and secondary cache size is typically write-through in which main memory is updated in concert with cache for cache writes.It is observed that multilevel on-chip caches improve system performance. High-performance architectures use small on-chip Level1 (L1) writethrough caches complemented by off-chip level2 (L2) variable sized writeback caches. But at higher clock rates, off-chip caches can't keep up with processor speeds; latency and cache throughput compromise performance. In system running at 200 MHz many clock cycles are used to bring data from the L2 cache. One solution is to bring larger, though somewhat slower, L1 caches on-chip. Alternatively, fast and small on-chip L1 caches working with larger higher throughput L2 caches; also on chip; can reduce average load latency. An L2 cache can provide data to the L1 cache and processor execution pipelines at rate exceeding 4 Gbytes/s. Moreover, using a dissimilar cache organisation for each cache level raises the probability that the information in the cache is correct and up-to-date.
Two-level on-chip cache hierarchies offer sveral advantages. If the L1 caches relatively small, a dual ported cache can be implemented with little overall die penalties, and two cache loads can be issued and executed at one time. In addition to accelerating performance, dual-issue cache designs eliminate the gate delays associated with a single large L1 cache.
Today's research has shown that even smaller objects might serve as storage
devices thus giving the idea of biological memory.The data stored in this way can be stable for five years.
Practically, researchers believe that Holographic data storage system in
which thousands of pages (blocks of data), each containing million bits,
can be stored within the volume of a sugar cube, have a storage capacity of
10 GB per cubic centimeter.This figure is still very impressive compared to
today's magenetic storage densities, which are around 100 Kb per square
centimeter (not including the derive mechanism).
At this density a block of optical media roughly the size of a deck of
playing cards would be able to house a terabyte of data.Because such
system can have no moving parts and its pages are accessed in parallel,
it is estimated that data throughput on such system can hit 1 Gbps or
higher. In holographic recording applications, longer interaction lengths
imply increased angular selectivity and also higher data storage capacity
. These advantages are in addition to the ability to synthesize a much
larger cross sectional area then is currently attainable using bulk materials.
Holostore leverages the imaging properties of light and its ability to launched. The reading out of images instead of single bits serially provides a tremendous improvement in the bandwidth. The ability for light to be launched through space and deflected easily will eliminate the need for rotation of the medium. The capability of coherent light to interfere and to form holograms provides a convenient way to address a storage medium in three dimensions, while only scanning the beams in two dimensions.
Holography records the information from a three-dimensional object in such
a way that a three dimensional image may subsequently be constructed.
Holographic memory uses lasers for both reading and
writing the blocks of data into the photosensetive material. A digital
hologram is formed by recording the interference pattern between a
discretely modulated coherent wave front and a reference beam on a
photosensitive material.
Despite decades of research in holographic-storage materials, iron
dopped lithium niobate is still the medium of choice for all
demonstrations of holographic-storage system.Despite its well known
shortcomings, such as destructive readout of data and relatively low
sensitivity, its the only material that currently has the optical
quality that is critical for a system application.
A difficulty with the holostore technology had been the destructive readout.
The re-elluminated reference beam (i.e the read beam, see How holographic
storage system works), used to retrieve the recorded information, also exites
the donor electrons and disturbs the equilibrium of the space charge field
in a manner that produces a gradual erasure of the recording. In the past,
this has limited the number of reads that can be made before the signal-to
-noise ratio becomes too low. Morover, wrtes in the same fashion can
degrade previous writes in the same region of the medium. This restricts the
ability to use the three-dimensional capacity of a photorefractive for
recording angle-multiplexed holograms.
The other challenge has been the geometry of the crystal medium. It is
difficult to grow large crystals of good optical quality and of limited size.
As a consequence of this problem, widespread application of the bulk
photorefrective technology has not been occured, despite an initial
surge of development in the decade of the 1970s.
The molecule in question here is the protein called bacteriorhodopsin
. Its photocycle, the sequence of structural changes, a molecule undergoes in reaction to light, makes it an ideal AND data storage gate, or flip-flop. According to the today's research, the bR (where the state is
0) and the Q (where the state is 1) intermediates are both stable for many
years.
A prototype memory system has built where bacteriorhodopsin stores data in a 3-D matrix. The matrix was build by placing the protein into a cuvette (a transparent vessel) filled with a polyacrylamide gel. The cuvette is oblong and 1 by 1 by 2 inches in size. The protein, which is in the bR state, gets fixed in place by the polymerisation of the gel. A battery of kypton lasers and a charge-injection device (CID) array surround the cuvette and are used to read and write data.
While a molecule changes states within microseconds, the combined steps to
read or write operation take about 10 milliseconds. However like the holographic storage, this device obtains data pages in parallel, so a 10 Mbps is possible. This speed is similar to to that of slow semiconductor memory.
The reason for considering the moleculer memory is that it is protein based and therfore is inexpensive to produce in quantity. Secondly, the system has ability to operate over a wider range of temperatures than semiconductor memory.
The heart of a biocontroller is a three state process involving:
This type of system is ideal for person with physical disabilities who do not have access to computers through the traditional manual inputs.
VR is a good gimmick with lots of applications such as point-
of-sale display, trade show booths and other advertising , or
publicity activities , where VR is an end it self -the novelty
of the medium sells the message. You probably won’t buy a
house solely on the strength of a VR walk-through, but the
novelty of it might tempt you to go view the real thing.
Some of the equipment used in VR are HMD’s (head mounted
display), data gloves and Wands.
These will be discussd below.
Perhaps the most important component in a VR system is
the enigmatic head mounted display (HMD). Since its invention
in the 1960s , HMD design and construction has surfaced in
several different approaches, each improving the realism
and/or price point of previous models. The most basic display
requirements for a HMD are two display screens and a set of
optics. The display screens present the information sent by
the computer and optics (in the form of lenses) permit the
user to focus on images that rest about 4 inches from their
face .
There are several variations in the type of HMDs available.
LCDs (liquid crystal displays), the kind in portable video
games, are most popular choice in todays design offering
colour output and a reasonable price tag. Unfortunately
today’s LCD screens don't deliver much in the way of
resolution, a problem amplified by the HMDs magnifying
optics. Typical LCD resolution stands at 440*240 split between
reed , blue and green,. This provides developers with a total
pixel count of 35,000. In comparison to standard VGA display
of more than 300,000 pixels, its easy to see why designers are
frustrated by limits of LCD screens.
The CRT (cathode ray tube ) alternative eliminates much of the
resolution problem, but creates new problems of both price ,
colour and size. CRT’s are small enough to be implemented in
to a HMD , but are not a common in other industrial
applications so finding them at a reasonable price can be
difficult. Another problem is the lack of a small display
CRT display unit that can deliver a full colour display.
Finally fitting one of these to into a HMD require complex
optical set-ups and electromagnetic shielding, both of which
add weight and size to the final design. There a ways to
address all of these problems, but the added expense has kept
such devices in the hands of the military or major
corporations and beyond the reach of the layman.
In order to achieve a true sense of realism, VR devices should
enable the user to interact with his or her world in a manner
as close as possible to the way they interact with the real
world. Even with the best stereoscopic displays a HMD wearer
is forced to use a keyboard , which constantly reminds one of
a simulation. Motion tracking eliminates this problem by
giving the computer a constant update on the orientation of
the headset. As the display itself, there are a few different
ways to achieve this particular goal.
Electromagnetic coils are currently the best way to achieve a
cheap and effective form of tracking. In which two sets of
three-wires coils have a small electric charge applied to them
in sequence, creating magnetic field measure by another
device. The chief advantage of EM tracking is that it is
small, and for most part inexpensive (nothing in VR is really,
cheap yet). The biggest disadvantage is the tendency of the
system to pick up outside signals from televisions and
monitors in the area. Ultrasonic trackers and mechanical
tracking (the use of chord or wires that move with the head)
are alternatives , but are more suited to specialist needs
than mass-market application.
Wands are much like the joysticks most videogamers are
familiar with, except they do it without a base. They are available in
almost every shape imaginable, they contain variations of the
same tracking equipment found in HMD’s with a couple of
buttons tacked on. Each axis of motion (pitch, roll and yaw)
is measured, and sent back to the computer which translates
the signal into motion, action, or whatever the programmer can
dream up. Wands are relatively cheap ,very effective and
perhaps most importantly they are easy for a new user to
understand and operate.
Spaceballs are the most familiar of the isometric family of
control units. These devices look like a round ball set into a
base, and use optical or mechanical sensors to detect any
force on the ball. If the user pulls the ball straight up,
twists it, or rolls it back and forth, an appropriate signal
is sent back to the computer. Other isometric units often look
like joysticks or plates, and are most useful for streamlined
motion control.
Gloves have been a part of VR history since the early ‘80s.
More than any other type of control, gloves enable users to
interact with the computer world in a way perfectly mimicking
reality.
While most input devices offer one, two or three degrees of freedom, the glove
is unique in that it offers multiple degrees of freedom for each finger and
hand as well. This permits the user to communicate to the computer , a far
richer picture of his or her intentions than most other devices.
A glove is generally quite lightweight , with flexible sensors which accurately
and repeatedly measure the position and movements of the fingers and wrist.
Pressure sensors on the gloves palms measure occurring during object grasping.
Gloves can have a varying number of sensors depending the application they are
used for. Most common models have in-between 12 and 24 sensors. These sensors
are bend sensitive and their resistance varies linearly with the bend. Sensors
are extremely thin and flexible and provide an undetectable resistance to
bending. Since the sensors exhibit low sensitivity to their positions over
finger joint and to the joint radii of curvature, gloves provide high quality
measurements over a wide range.
Glove variations include ones with open finger tips, which enable users to type,
write and grab objects while wearing the glove.
The glove is connected to the computer via an instrumentation unit. This
instrumentation unit is connected to the computer by a serial cable. The
instrumentation unit provides a variety of convenient functions and features,
including time-stamp , dataglove status , external sampling synchronisation and
analogue sensor outputs.
Software is necessary or calibration of the glove and is used to display event
virtually.
Exoskeletons are also employed to simulate the
resistance of objects in a virtual world. An exoskeleton is
basically a robotic arm strapped to a person. At the
University of Utah, researchers have developed a robotic arm
which has 10 degrees of freedom. The robot continuously
updates the force to each of its ten joints, and can make it
appear that the 50 pound arm is weightless. "However, when
the operator touches something, the virtual forces become
actual forces felt through the exoskeleton" (Lane and
Smith). This would make the operator's arm stop when it hit
a virtual wall or feel the weight of a virtual object.
“Tactile feedback” is another glove innovation that’s been
pursued for some time. In VR simulation, though users can
access a glove to pick up objects, his or her sense of touch
will not give then the signal that they would receive in the
real world, a reminder that they’re in a simulation. First
pioneered by the advanced Robotics Research centre in England,
the first tactile feedback system to achieve a believable
tactile response used small air bladders inside the glove that
could return pressure in 20 different areas. Because the
bladders currently tend to cause glove calibration problems,
this method is still being perfected. When working correctly,
a user can reach out in a virtual world to touch an object and
actually feel the physical response of the object on his or
her skin. Other research in tactile feedback is being done
with the use of small vibrating coils. Although these coils
aren’t as realistic in feel as the air bladders , these units
are far cheaper to make . and to deliver some believable
feedback.
Speculation about the future of VR has brought up very serious
questions and others aren’t so serious . Questions about
health risks on VR, physiological effects of long term
isolation and so on.
On the most basic level, the long term effects of having two
monitors strapped inches from your eyes is in question
.Headsets weighing more than the head can comfortably support
often intensifies the effect, as do lag times between the
motion of the head and the resultant display. Eye strain is
also a negative side effect.
Also several studies have been run to attempt to determine the
long term effects of isolation on human physiology. The
question is if a person spend most of his time in a fantasy
world, will he be able to relate to people in the real world
when they emerge?
Some people think that the HMD, today’s symbol of virtual
reality may be obsolete before it is even properly built: HDTV
(high definition television ) will deliver huge pictures with
resolution so perfect that we will be able to see every eye
lash on a persons face.
Another new possibility (being worked on , is a system that
fires low powered lasers through the eye , actually painting
the image onto the retina . Resolution-wise this technique has
potential to deliver picture that simulate every rod and cone
so as to create an image that is ‘perfect’.
Tomorrow's doctors may pilot small remote controlled drones
capable of repairing internal injuries without extensive
surgery. Doctors use head sets to examine areas of the body
from the inside, and glove to control the actions of tiny
surgical units. Data suits can be used to train athletes and
are used in rehabilitation.
Hey people might even be able to shake hands in cyberspace.
The next time you stare at your computer you might be using
your eye as an input device. Consider this , the instant you
eyes stop moving and you are staring at a menu item or icon -
approximately 250 ms later , the application you were staring
at opens up like magic. Is it the computer reading your mind
or playing a guessing game with you? We will soon find out.
A new generation of research using eye-tracking hardware and
on -screen graphical symbols has begun. Easy applications
such as hands-free command input for fighter pilots ,
historically have received the most research attention.
However now eye- tracking developers are tackling the more
difficult work -making the technology efficient for general
business applications. Some benefits of eye-tracking are
obvious. Lab studies have shown that the technology lets you
work about 25 percent faster than with a traditional mouse.
Hands-free input also means no carpal-tunnel problems or
tendinitis. Once more hands free input may be the only method
for disabled people to interact with the computer. Some People
think eye-tracking is ushering a new era of non-command
interface that engage computers and humans in a continuos
dialogue. Rather than waiting for you to issue discrete
commands, these interfaces monitor body language to anticipate
your next move. Eye-tracking can be especially important when
you are browsing for information, whether its application
icons scattered across the screen or hotlinks scattered
throughout the document and performing precise tasks such as
cutting and pasting.
Although commercial eye-tracking hardware has dropped from
about $250,000 to $20,000 in the last eight years, the
hardware can be unstable. A system that works well with one
person might stumble when someone else sits at the computer.
This is probably because people with larger pupils have the
greatest success with eye tracking. (1The pupil is the command
centre of eye-tracking systems. ).
Eye-tracking hardware shines an infrared beam into the users
eye to illuminate the pupil so that a custom video camera
mounted on the computer can record the pupil image. Image
processing hardware digitises the pupil image and turn eye
movements into horizontal, vertical and depth cordinates.
Unlike earlier system where the head had to be kept still for
the system to work , current systems factor out head movement
by reading a non-pupil reference, typically a point of light
reflected from the cornea.
Will eye tracking dominate the desktops in time to come. I
guess we will have to wait and see. It is widely believed that
our eyes will be one component of future “multimodal input”
systems. When those systems come we’ll be gazing, talking and
pointing at our computers to give them commands. And in return
our computers will become more adept at deciphering how our
bodies say what's on our mind.
2.4 Buses
Current state of the art bus technology include universal serial buses(USB),
serial storage architecture (SSA) and fiber channel arbitrated loop(FC-AL). FC-AL is
probably the most modern and efficientof the three mentioned above. It provides
a substantial bandwidth improvements over SCSI. FC-Al speeds are around 100MBps.
However several products are dual ported and have a total bandwidth of 200MBps. Dual porting also
increase fault tolerance by offering redundant access to each device. Overall
the FC-AL is at least twice as fast as the fastest SCSI available. It has 3
important features
1)It enables hot plugging
FC-Al primarily targets high endserver applications, where you need all the
performance you can get. FC-Al's fault tolerance is an added advantage.
It hasn't exactly revolutionised the desk tops for several reasons. Single user
desktop applications running on standard Intel boxes might not be able to take
advantage of the FC-AL speed. Fault tolerance is not critical for the desktop
market. It is also quite expensive to implement.
2)Many host computers support FC-Al for host to host interconnections.
3)FC-Al supports cables up to 30m long with coaxial cables and up to 10m with
optical fibre.
3. Future of Computing
3.1 Optical & Biological Possibilities
Although we are living in the "computer age", the full implications of computational ideas have not been realized in our century.
The capacities of today's mass storage devices cannot satisfy the demands
of new processes which will be developed near future.To achieve a full
scale opticle computing environment (with large computing power), it
is necessary to have memories with rapid access time and large storage
capacity.To meet these needs holographic memories have emergrd.
3.1.1 Holostore Technology
A volume holographic storage (holostore) devive is a page oriented devive
that writes and reads data in an optical form.The holography technology
achieves the necessary high storage densities as well as fast access times.
This capability occurs because a holographic image, or hologram, encodes
a large block of data as a single entity in a single write operation.
Conversely, the process of reading a hologram retrieves the entire data
block simultaneously.
3.1.2 Moleculer Memory
With the advances in Moleculer electronics, it is possible to implement
a prototype memory subsystem that uses molecules to store digital bits.
3.1.3 BioControl Systems and BioSignal Processing
BioControl Systems (BCS) has created a new human computer interface(HCI) using state-of-the real-time biosignal processing. This technology enables an individual (including physically disabled individuals) to control a computer using eye movement, muscle tension and gestures, and even thought patterns.
This configuration allows flexibility in mapping biosignal inputs to output code and also allows for specific applications to be created in software without the need for specialized transducers on the body. The device has eight input channels with programmable gains and filters that can be configured to process virtually any bioelectric signal. An eye controlled joystick and mouse has been developed that allows users to control graphical objects with eye movements. The mouse applications give users the ability to do "hands free" word processing or sue other menu driven software.
3.2.1 HMD (head mounted display)
Motion-tracker
3.2.2 Wand
Spaceballs
3.2.3 Gloves
Exoskeletons
Tactile feedback of the glove
Problems
What may happen as we proceed deeper into the Future ?
3.2.4 Eyecons
How it works?
References
Ali Yaqoob
Last modified: Tue Jun 10 07:25:06 PDT 1997