Neena Imam

1
Presented at RAMS Faculty Workshop Oak Ridge,
TN December 10, 2007
Algorithm to Ultra-fast Signal Processing
Highlights of Selected Complex Systems Research
Activities

• Neena Imam
• Complex Systems
• Computer Science and Mathematics Division
• OAK RIDGE NATIONAL LABORATORY

2
Outline

• Introduction
• acknowledgments collaborators
• overview of Complex Systems
• Research activities
• missile tracking and interception
• hyperspectral sensors
• sonar signal processing
• quantum devices
• Future directions and contacts for collaboration
• collaboration topics
• Complex Systems contact points

3
Acknowledgements
for activities presented hereafter

• Collaborators
• Jacob Barhen
  ORNL / Complex Systems (Group Leader)
• Travis Humble ORNL / Complex
  Systems
• Jeffery Vetter ORNL / Future
  Technologies
• Aeromet Corporation Tulsa, OK
• Thomas Gaylord Georgia Tech
• Eustace Dereniak U. Arizona
• Albert Wynn, Deirdre Johnson students, Research
  Alliance for Mathematics and Science
• Technology Sponsors
• Missile Defense Agency
• Naval Sea Systems Command
• Office of Naval Research

4
Complex Systems Overview
Mission Innovative Technology in Support of DOE
DOD Theory Computation
Experiments

• Research topics
• Missile defense C2BMC (tracking and
  discrimination), NATO(ALTBD), flash
  hyperspectral imaging.
• Modeling and Simulation Sensitivity and
  uncertainty analysis of complex nonlinear
  models, global optimization.
• Laser arrays directed energy, ultraweak signal
  detection, terahertz sources, underwater
  communications, SNS laser stripping.
• Terascale embedded computing emerging multicore
  processors for real-time signal processing
  applications (CELL, Optical Processor, ).
• Anti-submarine warfare ultra-sensitive
  detection, sensor networks, advanced
  computational architectures, Doppler-sensitive
  waveforms.
• Quantum optics cryptography, quantum
  teleportation (remote sensing).
• Computer Science UltraScience network.
• Intelligent Systems neural networks,
  multisensor fusion, robotics.
• Materials Science control of friction at micro
  and nanoscale.

Sponsors DOD(DARPA, MDA, ONR, NAVSEA ),
DOE(SC), IC (CIA, IARPA, NSA), NASA, NSF
5
TARGET TRACKING AND DISCRIMINATION
6
MDA's HALO-II/AIRS Project

• Independent Verification and Validation (IVV)
  of software.
• Improved tracking algorithm development.
• Sensitivity analysis of system modules using
  Automatic Differentiation (AD).

7
Motivation For HALO-II/AIRS

• Meet MDA TE Requirements
• Sensor / Technology Testbed

Orbital Signatures
Exo-AtmosphericTarget Characterization
Chemical Releases
Counter- measureSignatures
Vehicle Separation
Plume Signatures
Target Signatures
Kill Assessment orMiss Distance
TrajectoryReconstruction
Booster Tracks
Failure Diagnostics
Photo documentation
FOR
Interceptor Performance
Flash Radiometry
8
HALO-II System Overview

• Five Subsystems. Sensors installed in
• aerodynamic pod.
• In-Pod
• Pointing
• Acquisition
• Tracking
• In-Cabin
• Real time processor
• Surveillance processor

In-Pod
In-Cabin
highest level view
9
Sensitivity and Uncertainty AnalysisMotivation
How much confidence should be placed in decisions
obtained on the basis of predictions from complex
mathematical and / or physical models embedded in
complex code systems?

• Uncertainties
• input data
• outputs
• model parameters
• sensor measurements

Code B

• For example, modeling of battlespace threat
  signatures encompasses a large set of varied
  phenomenologies
• importance of accurate threat signature
  discrimination precludes confidence analysis
  based solely on parameters and model features
  selected by engineering judgment.

10
Sensitivity and Uncertainty AnalysisObjective
Recognized need for computational tools that
explicitly account for model sensitivities and
data uncertainties. The design of complex
multisensor-based targetdetection / tracking
architectures illustrates typical application.

• The methodology has two primary goals
• determine confidence limits of predictions by
  large code systems
• consistently combine sensor measurements with
  computational results
• obtain best estimates of model parameters
• reduce uncertainties in estimates

N. Imam and J. Barhen, Reduction of
uncertainties in the USNO astronomical refraction
code using sensitivities generated by Automatic
Differentiation, 2004 International Conference
on Automatic Differentiation (7/04), Chicago, IL.
11
ORNL Developed Improved NOGA Tracker
Simulation ResultsElevation and Elevation
Uncertainty Sensor Data vs HALO prediction
NOGA is an ORNL developed method that produces
best estimates for quantities of interest by
explicitly incorporating uncertainties in the
estimation process. It involves a fast, nonlinear
Lagrange optimization. The tracking implemented
in conjunction with NOGA is based on a second
order auto regression.
N. Imam, J. Barhen, and C. W. Glover,
Performance evaluation of time-weighted
backvalues least squares vs. NOGA track
estimators via sensor data fusion and track
fusion for small target detection applications,
Proc. of SPIE, Signal and Data Processing of
Small Targets, vol. 5913, pp. 59130Z1- 59130Z1,
2005.
12
Sensitivity Analysis of the Airborne Pointing
System Module

• APS drives the sensors. Calibrates using USNO
  astronomical refraction code.

• Astronomical Refraction Observer in earths
  atmosphere,
• object outside. USNO code uses numerical
  integration.

The real part of the atmospheric index of
refraction is a nonlinear function of pressure,
temperature, elevation, humidity, and
wavelength. Therefore, light propagating in the
vertical direction is bent towards lower altitude.

• ORNL devised experiments to improve APS
  performance after sensitivity analysis was
  completed. The sensitivity and uncertainty
  analysis highlighted the approximations/limitation
  s inherent in this model and aid in the design of
  more accurate refraction algorithms.

calculated response sensitivities input parameters
13
SONAR SIGNAL PROCESSING
14
Wideband Sonar Signal Processing

• For wideband signals, the effect of target
  velocity is no longer approximated as a simple
  "shift" in frequency.
• Doppler effect a compression/stretching of the
  transmitted pulse.
• Wideband Ambiguity Function (WAF) a function of
  time delay t and Doppler compression factor h.
• Doppler Cross Power Spectrum (DCPS) forms a
  Fourier pair with the ambiguity function and can
  be used to calculate the ambiguity function and
  the Q function 1, 2

1. R. A. Altes, "Some invariance properties of
the wideband ambiguity function," J. Acoust. Soc.
Am. 53, pp. 1154-1160, 1973. 2. E. J. Kelly and
R. P. Wishner, "Matched filter theory for high
velocity accelerating targets," IEEE Trans. Mil.
Electron. MIL 9, pp. 59-69, 1965.
15
Wideband Ambiguity Function

• For a low Q function, and hence a high
  reverberation processing, it is necessary to
  minimize the area under the square of the modulus
  of the DCPS along a line of constant Doppler
  scaling 1.
• spread the energy of the transmitted pulse over a
  broad bandwidth
• CW signal can use a very narrow bandwidth to
  achieve low Q but compromises parameter
  estimation
• use of Comb spectrum, SFM or LFM signals

here w(t) is the window function B bandwidth
1.T. Collins and P. Atkins, "Doppler-sensitive
active sonar pulse designs for reverberation
processing," IEE Proc. Radar Sonar Navig.
145, 347-353 , 1998.
16
Ambiguity Functions of DSW
17
Matched Filtering for Active Sonar Processing
A synthetic echo is generated for a particular
target range and velocity. The echo signal is
correlated with a bank of replicas. Spectral
techniques are used. The correlation with the
highest magnitude provides an estimate of the
Doppler velocity bin. The location of the maximum
within that correlation yields the time delay of
the echo, and thus provides an estimate of the
range.
18
Matched Filtering for Active Sonar Processing

• SFM pulse of fc1200 Hz
• Bandwidth B 400 Hz
• Pulse duration 1 s
• Modulation frequency 5 Hz
• Sonar sampling rate fs 5000Hz
• FFT length 80K

• Target
• assumed range 3Km
• assumed velocity - 5m/s (bin1)
• 32 matched filter bank.
• Result
• output of the first filter has the closest match
  to the received signal.
• Time delay 4 seconds thus, estimated target
  range 3 Km.

19
The EnLight TM Prototype Optical Core Processor

• Full matrix ( 256 x 256 ) - vector multiplication
  per single clock cycle
• Fixed point architecture, 8-bit native accuracy
  per clock cycle
• Enhanced by on node FPGA-based processing and
  control
• Demonstrated accuracy and performance in complex
  signal processing tasks
• Developed by Israeli startup

EnLight 64? demonstrator

• Power dissipation (at 8000 GOPS throughput)
• EnLight 40 W (single board)
• DSP solution 2.79 kW 62 boards, 16 DSPs
  (TMS320C64x) per board

Information provided by Lenslet, Inc
20
Matched filter calculation on EnLight-64?
hardware
Performance Comparison
-30
MATLABAlpha
MATLABAlpha
-35
-40
Output of filter 1, dB
-45
-50
-55
2000
2600
4000
3200
3400
3600
3800
2800
3000
2400
2200
Range (meters)

• Speed-up factor per processor
• E_64? 6,826 ? 2 gt 13,000 actual hardware
• E_256 56,624 ? 2 gt 113,000 emulator

• Computation parameters
• FFTs 80K complex samples ? number of filter
  banks
• 33 filter banks 32 Doppler cells,
  1 target echo

21
HYPERSPECTRAL IMAGE PROCESSING
22
Hyperspectral SensorComputer Tomography Imaging
Spectrometer (CTIS)

• CTIS Simultaneously acquires spectral
  information from every position element within a
  2-D FOV with high spatial and spectral
  resolution.
• CTIS is being developed at Optical Detection Lab
  of U. Arizona by Eustace Dereniak et. al.

Objective is to collect a set of registered,
spectrally contiguous images of a scenes
spatial-radiation distribution within the
shortest possible data collection time
23
CTIS Instrumentation at U. Arizona
24
CTIS Principle
Mapping of signal from the object cube to the
focal plane array
25
CTIS Code Acceleration

• Computationally demanding
• Convergence issues
• An example reconstruction
• 5 sec for each iteration for a 0.1 micrometer
  spectral sampling interval (3-5 m region) and
  46X46 spatial sampling. Total of 46X26X21
  sampling. 10 iterations needed for
  convergence. 1/3 hour computation time for each
  frame.

Algorithms must be developed for less
computational time and better convergence

• Improved algorithm employing conjugate gradient
  method
• Parallel programming for CELL Broadband Engine
  (CBA) multicore processor
• Reconfigurable computing via FPGAs

26
IBM Cell Multicore Device

• Research Centers contributing
• IBM USA
• Austin, TX (lead, STIDC)
• Almaden, CA
• Raleigh, NC
• Rochester, MN
• Yorktown Heights, NY
• IBM Germany
• Boeblingen
• IBM Israel
• Haifa
• IBM Japan
• Yasu
• IBM India
• Bangalore

• CELL Broadband Engine Architecture (CBEA)
  jointly developed by Sony, Toshiba and IBM
• Took 5 years, over 400 Million dollars, and
  hundreds of engineers
• New design relies on heterogeneous multicore
  architecture
• abandons mechanisms such as cache hierarchies,
  speculative execution, etc
• based on fast local memories and powerful DMA
  engines

27
Mapping Communications to SPEs

• Original single-threaded program performs many
• computation stages on data.
• How to map to SPEs?

• Each SPE contains all computation stages. Split
  up data and send to SPEs.

• Map computation stages to different SPEs.
• Use DMA to transfer intermediate results from
  SPE to SPE in pipeline fashion.

28
Overlapping DMA and Computation

• We are currently doing this

• We can use pipelining to achieve
  communication-computation concurrency.
• Start DMA for next piece of data while
  processing current piece.

29
Reconfigurable Computing via FPGAs

• The emergence of high capacity reconfigurable
  devices has ignited a revolution in
  general-purpose processing.
• It is now possible to tailor and dedicate
  functional units and interconnects to take
  advantage of application dependent dataflow.
• Early research in this area of reconfigurable
  computing has shown encouraging results in a
  number of areas including signal processing,
  achieving 10-100x computational density and
  reduced latency over more conventional processor
  solutions.
• FPGA, short for Field-Programmable Gate Array, is
  a type of logic chip that can be programmed.
• An FPGA is similar to a PLD, but whereas PLDs
  are generally limited to hundreds of gates, FPGAs
  support thousands of gates.

SPECT Laboratory is involved in the development
and demonstration of latest generation FPGA
computing applications.
30
Xilinx XtremeDSPTM FPGA Hardware

• 500 MHz Clocking.
• Multi-Gigabit Serial I/O.
• 256 GMACS Digital Signal Processing.
• 450 MHz PowerPC Processors with H/W Acceleration
  .
• Highest Logic Integration.
• 200,000 Logic Cells.
• Reduced Power Consumption.
• Achieve performance goals while staying within
  your power budget.

VIRTEX-4 XtremeDSPTM Development Board
The Xilinx XtremeDSP initiative helps develop
tailored high performance DSP solutions for
aerospace and naval defense, digital
communications, and imaging applications.
31
FPGA Signal Processing Station at SPECT Laboratory

• Pegasus Demo Board with SPARTAN-2
• Digilent VIRTEX-2 Development board
• VIRTEX-4 XtremeDSPTM Development Board

32
QUANTUM HETEROSTRUCTURES
33
Quantum Heterostructures

• Heterostructures consist of alternating layers
  of semiconductor materials of similar lattice
  constants.
• Quantum confinement alters the electronic band
  structure.
• Electron potential can be tailored by
  appropriate choice of materials.

• Electronic energy levels are discretized
  resulting from one-dimensional confinement
  potential of semiconductor heterostructures.
• The levels are broadened into subbands due to
  the in-plane momentum of carriers.

34
Intersubband Lasers and Photodetectors
Quantum Well Infrared Photodetector (QWIP)
Intersubband Laser
Bound to continuum transition

• Voltage tunable (7 mm - 9 mm).
• Dl/l 10-3.
• Multicolor detectors.

• l 3 mm - 11 mm .
• 300 K pulsed, CW up to 110 K.
• Dual wavelength (8 mm, 10 mm) lasers.

35
Applications of Intersubband Devices

• Medical treatment

• Wireless infrared networks

• Computer networking

• Remote sensing
• Earth science monitoring

36
Quantum Well Infrared Photodetector (QWIP)

• Voltage tunable.
• Dl/l 10-3.
• Multicolor detectors.

• Bound eigen-states have real energies.
• Types 1 and 2 quasibound states have complex
  energies.

• Apply transfer matrix method to structure
• to find equivalent matrix M.
• Use APM to find the zeros of the complex
  function Det(M)0 to determine the eigen-states

Argument Principle Method (APM)
37
QWIPs for Multicolor Infrared Detection

• Using bandgap engineering it is possible to
  extend the functionality of a
• QWIP for multicolor detection.
• Multispectral applications may be very useful in
  spectral analysis of Infrared sources and target
  discrimination.
• In one possible configuration, several
  conventional QWIP structures with
• different selectivity are stacked together.
• Use different transitions within the same
  structure. Symmetric and asymmetric wells have
  been used.

Grave et al., Appl. Phys. Lett. 60, 2362 (1992).
Kheng et al., Appl. Phys. Lett. 61, 666 (1992).
Martinet et al., Appl. Phys. Lett. 61, 246 (1992).
38
Design Methodology of An Optimized QWIP

• Eigen-state determination using APM.
• Dipole matrix (absorption strength) calculation.
• Self Consistent Solution Two factors contribute
  to carrier potential energy.
• Poissons equation and Schroedingers equation
  must be solved iteratively until convergence is
  achieved.
• Cost Function Formulation and Iterative
  Optimization simulated annealing, genetic
  algorithm etc.

39
Absorption Spectrum of Bicolor Equal-Absorption-Pe
ak QWIP Structure at Room Temperature

• MCT detector
• 90, 000 scans

• DE12 134 meV, l12 9.25 mm.
• DE13 193.4 meV, l13 6.4 mm.
• R 0.71.

Imam et al., IEEE J. Quantum Electron. 39, pp.
468-477, 2003

• Sharp, well resolved peaks, Lorentzian in
  Lineshape, no other peaks present.
• The absorption spectrum is very high quality and
  has little noise due to large number of scans
  taken .

40
Current and Future Directions in Quantum
Heterostructure Devices

• Multi-wavelength detectors
• Hyperspectral sensors
• Room-temperature devices
• Less costly devices
• Improved device modeling and simulation

Bandgap Engineering is the key!!
Imam et. al. Superlatt. Microstruct., vol. 28,
pp. 11-28, July 2000. Imam et. al. Superlatt.
Microstruct., vol. 29, pp. 41-425, June 2001
. Imam et. al. Superlatt. Microstruct., vol. 30,
pp. 28-43, Aug. 2001. Imam et. al. Superlatt.
Microstruct., vol. 32, pp. 1-9, 2002. Imam et
al., IEEE J. Quantum Electron. Vol. 39, pp.
468-477, 2003.
41
Examples of Possible Collaboration Topics

• Algorithms for Vectorized Fourier Transforms and
  Implementation on Multicore Processors.
• Digital Signal Processing Design and FPGA
  Implementation.
• Quantum Well/Dot Device Modeling, Simulation, and
  Fabrication.
• Tracking Algorithm Development.

42
Contacts
Center for Engineering Science Advanced Research
(CESAR) Computer Science and Mathematics
Division Oak Ridge National Laboratory
Neena Imam Research and Development
Staff Phone [phone] Fax [phone] E-mail
imamn_at_ornl.gov Jacob Barhen Group
Leader Phone [phone] Fax [phone] E-mai
l barhenj_at_ornl.gov 1 Bethel Road Bldg 5600, MS
6016 Oak Ridge, TN 37831-6016 USA
Patty Boyd Administration Phone [phone] Fax
865-574-0405 E-mail boydpa_at_ornl.gov