Transcript Document

Analog-Faster-Cheaper-Better
An Optical Signal Processing View
Terry Turpin
Chief Scientist Essex Corporation
[email protected]
Facts
• The “The Universe” is analog
• Human technology is still mostly analog
– (did you ever see a digital bicycle)
• Digital has dominated the information processing and
communications world for more than three decades
• Analog processing has been ignored by educational
institutions
• There are at least two generations of scientists and engineers
that have never learned analog processing or communications
technology
• Analog optical processing in the past as been a small but
persistent exception
• Optical communications and analog optical processing are
merging the same way that digital processing and digital
communications did in the past
Processing Overload
World of Analog Signals
Spread
Spectrum
A
Fiber Optic
to Digital Stream
Microwave
D
Land Lines
Radio 3,488 Tbs
Television 68,955 Tbs
Telephone 17,300,000 Tbs
Internet 532,897 Tbs
17,905,340 Tbs!!
(2002 Data!)
The Information Superiority Problem
So much information…
so little time to process it
Processing power is the key to
superiority in a world market
Summary of electronic information flows of
new information in 2002 in terabytes
… 17.7 exabytes each year, and growing
“Era of Tera”*
… a Digital Perspective
*Pat Gelsinger, CTO Intel
(Keynote address at Intel Developer Forum Feb 2004)
Digital Dilemma over Power
“Power density is increasing at a rate that implies that
tens of thousands of watts per centimeter (w/cm2) will be
needed to scale the performance of Pentium processor
architecture over the next several years. But that would
produce more heat than the surface of the Sun…”*
*Pat Gelsinger, CTO Intel
(Keynote address at Intel Developer Forum Feb 2004)
… Begins the Age of Optical
*
Optical Processors
Analog Optical Processors excel at
- Images
- Signals
- Correlations
*Pat Gelsinger, CTO Intel
(Keynote address at Intel Developer Forum Feb 2004)
References to Optical Processors added by Essex Corp.
Analog Optical Processing Overview
Optical Processing
Today
Future
Measured By
Faster
Yes
Yes
TIPS
Cheaper
Sometimes
Yes
Bytes/$
Cooler
Always
Always
Watts/cm2
Better
Sometimes
Sometimes
Performance/
Application
An Unclassified Success in
Size, Weight and Power
• Acousto-Optic Spectrometer AOS launched late
1998 on SWAS for a 2 year mission
• 4 channel 1400 point Fourier transform in real
time on a 1.4 GHz analog signal
• Compute power is 500 Gigaflops (Sustained) for
12 Watts electrical power
• Analog input eliminated the need for high speed
A/D converters
• Mission to study the chemical composition of
interstellar clouds
• SWAS would be impossible without the AOS
optical computer
Optical Processor/Computer?
… a machine
that performs
mathematical functions
with light rather than
electrons
Functions most frequently used
•Fourier Transform
(demultiplexing/multiplexing)
•Correlation (pattern detection)
•Data distribution and replication
Why go to Analog Optical Processors?
• Speed
Advantages
• Reduced size and power consumption
• OEO Overhead & Cost are Excessive
(optical - electrical - optical)
• Natural Fit: Optical Processing for
- Optical Communications
- Images
- Signals
- Correlations
• Typical improvement is a factor of 50000
Information on Light
• Information is carried by the complex-valued
property of light (spatial frequency, amplitude
and phase)
• When an information-carrying beam is passed
through a special lens or coating, or interfered
with another reference beam, light performs
mathematical functions
Massive Parallelism
• Operates simultaneously on an entire wave
front and more than one variable — e.g.,
direction, amplitude and phase
• Digital systems are serial in nature
• Example: A lens simultaneously acts on the
entire light beam
Computational Set
• Analog optics can perform mathematical
functions
–
–
–
–
–
–
add
copy
multiply
Fourier transforms
correlation
convolution
• Operates on one- and two-dimensional arrays of
numbers in parallel
• A single analog optical computer “instruction”
might require thousands or millions of individual
instructions for a conventional computer
Analog Optical Computing
Combines the best of both worlds:
precision of electronics with
massive computational power of light.
Smaller,
Lower Power,
Lighter Computers
Optical
Computational
Module
12 inches square
Vs.
Many Parallel
Electronic
Processors
Supercomputer power where
it can’t go now.
• Head of a missile
• UAV
• Mobile ICBM Defenders
• Satellites
Cutting Edge Elements
Materials
•Photonic Crystals
•Non-Linear Materials
•Silicon Germanium
•III-V & II-VI Materials Systems
New Components
•VCELS
•Optical Fiber
•Optical Amplifiers
•SOA
•EDFA
•Optical Correlators
•Optical Signal Processors
Technology
•Photon Echo
•Optical Tap Delay
•Solotons
Example: Analog Optical Encryption
• Digital Encryption
–
–
–
–
ATM at 10Gbps soon
No 40 Gbps on horizon
Protocol specific
Cost increases linearly with number of signals on a
fiber
• Analog Encryption
–
–
–
–
5000 Gbps on horizon (ESSEX Eclipse Module)
Potential for multi-band encryption (L,C, and S)
Protocol agnostic
Cost is market driven and grows slowly with
capacity on a fiber
– 100 Teraflops for less than 10 Watts of electrical
power
Hyperfine Analog Optical Encryptor
X Gbps
Phase Key
Control
Computer
Sub-channels
Point B
Phase Key
Control
Computer
Sub-channels
X Gbps Device
or
WDM
Devic
e
…
X Gbps
Hyperfine Device
Analog Decoding
“key”
Reflective Phase
Modulator Array
Analog Encoding
“key”
C Band Comms Device
Reflective Phase
Modulator Array
…
WDM
Devic
e
X Gbps Device
or
Hyperfine Device
Point A
C Band Comms Device
How Does it Work?
Input
Transmitted
Scrambled Photons
Recovered
Simulated Data
Terabit Security – Cost
Perspective
*
*
**
*Assumes that aggregate bandwidths
above 10 Gbps will be encrypted using
multiple 10 Gbps encryptor pairs – 1 pair
per wavelength
** Estimated
**
costs are based on a multiplexed
optical signal with aggregate bandwidth as
indicated, and single optical encryptor pair per
optical link
Additional Advantages
• Analog optical processing provides an alternate
approach to thinking about problems
• This alternate approach often leads to solutions
that are radically different and sometimes better
• For example, to implement continuous scale
change and Fourier transforms on data that has
not been sampled or digitized
• Enable solutions to problems that are thought to
be too complex to solve economically
• In supercomputing applications the improvement
is about a factor of 50000
Summary
• Analog is faster, cheaper and better
• Examples are
– Separating signal channels in frequency
– Optical Encryption
– Optical Communications
– Optical Signal Processing
• Analog is a key technology
• Analog optical technology will force analog electronics
because of the A/D conversion limitation
• In optical communications, analog encryption and
wavelength routing will provide growth at low cost per
terabit