Transcript AlexSoiguine04242009

Alexander Soiguine [soi’gin]
(Sasha – standard Russian nickname for Alexander)
• Physicist, Mathematician, Computer Modeling Scientist
• PhD in Applied Mathematics, MS in Mathematical
Physics
• 20+ years of hands-on software development in
computer simulation
• Former professor of Mathematics and Mechanics
• Enthusiast of naturally parallelizable simulation
algorithms
Pasadena CA, April 24, 2009
• Physical Department of St. Petersburg University, Russia.
• MS in Mathematical Physics.
Major Courses:
- Higher Mathematics
- General Physics
- Atomic Physics
- Theoretical Mechanics
- Thermodynamics and
Statistical Physics
- Optics
- Electrodynamics
- Quantum Mechanics
- Mathematical Hydrodynamics
- Methods of Mathematical Physics
- Advanced Mathematical Analysis
- Theory of Scattering
- Operator Theory
- Nonlinear Problems of Mathematical Physics
- Spectral Theory of Operators
- Generalized Functions
- Wave Propagation in Random Media
• PhD in Applied Mathematics - Numerical Methods for
Stochastically Disturbed PDEs (Definitions of stochastic
solutions, existence/uniqueness theorems, propagated
variance estimations)
Pasadena CA, April 24, 2009
• Professor of Mathematics and Mechanics at Naval Academy of
Russia (St. Petersburg) – top level institution preparing senior
officers for Russian Navy.
Taught:
- Linear Algebra
- Probability Theory
- Differential Equations
- Vector Analysis
- Numerical Methods
- Coding Theory
- Stochastic Processes
- Partial Differential Equations
- Control Theory
- Theory of Optimized Filtration and Estimation
- Equations of Mathematical Physics
- Electrodynamics
- Heat Transfer Processes
• Participated in multiple research projects, particularly:
- submarine nuclear power plant numerical simulation algorithms
- relativistic particle beam dynamics simulation (star war challenge)
- acoustic emission simulation in metal construction
• Beginning of object-oriented programming
Pasadena CA, April 24, 2009
• January 1997 – first USA job as software developer at
International Metrology Systems - a company producing coordinate
measuring machines. Wrote C/C++ code to geometric
surface/shape/solid modeling
Programming was mainly around:
-
implementation of Levenberg-Marquardt best fit approximation
method for predefined geometrical shapes and spline-type
curves/surfaces
-
creating effective iteration algorithms to accurate and in real-time
construction of tangents/normals to curves/surfaces
-
integration of Matra Datavision Cascade geometrical object
library into existing software tool
-
OpenInventor/VRML 3D visualization routines
-
implementation of interprocess dynamical data exchange
multithreading mechanism
Pasadena CA, April 24, 2009
• August 1999 – took offer from Cadence to work on math
simulation engine for integrated electronic circuits.
Architectural, data structure modifications, matrix solver and
integration algorithms on-fly switching optimization.
Programming issues:
-
Restructuring circuit matrix nodes data from
linked list of doubles type into single double values, rewriting
device load member functions
-
Rewriting class member function interfaces for compatibility with
NIST SparseLib package
-
Restructuring circuit matrix load bypass algorithms for new
architecture
-
Creating algorithms to optimize switching between different
matrix decomposition and integration schemes
The result:
Electronic circuit simulation speed increased up to 15 times;
simulation began to converge with circuits where it never did before.
Pasadena CA, April 24, 2009
• 2003 – 2004 – Principal Analyst/Engineer at JRM Technologies.
Software simulation of physical signatures appeared due to optical,
electromagnetic, thermal, chemical phenomena. Implemented both
on Windows and Linux platforms.
What was done:
-
Algorithms to transform stellar brightness data into irradiation
approaching earth surface
-
Algorithms to calculate spectra of a city site illumination based on
given distribution of light sources
-
Simulation algorithm of radiation propagation from missile plumes
-
Simulation algorithms of thermal irradiation from vehicles, tanks,
detection by phase antennas
-
Terrain scene material identification algorithms based on satellite
hypespectral images
Pasadena CA, April 24, 2009
• Since August 2004 – computational scientist at Tinsley. Two
“from a scratch” huge projects - software tool to
visualize/analyze/process optical surface data collected from
coordinate measuring machines and interferometers, and software
package to interactively control optics processing machines.
Some of the first tool numerical blocks:
-
Algorithms to convert scattered point Cartesian data into regular
grid surface data
-
Algorithms to convert interferometer phase data into regular grid
surface data
-
Algorithms for surface profile analysis
-
Algorithms to filter regular grid data through cutting off 2D
rectangular/circle frequency domains
Pasadena CA, April 24, 2009
-
Algorithms to approximate data through Zernike polynomials
-
Algorithms for polynomial, moving average, weighted Gaussian,
other types of smoothing
-
Image cropping, clipping, stitching algorithms
-
Algorithms to create regular surface grid from given aspheric
equation
-
Algorithms to best fit metrology data to analytically predefined
aspheric surface
-
Algorithms to define pixels height/slope statistics in areas close to
optics edges
-
Algorithms of manual and scenario automated pixel editing
-
Image algebra
Pasadena CA, April 24, 2009
The following slides are a few
screen shots of the software tool to
visualize/analyze/process optical
surface data collected from
coordinate measuring machines
and interferometers.
Live demonstration will be
available.
Pasadena CA, April 24, 2009
How it works. Main data
visualization/analysis box
Pasadena CA, April 24, 2009
Profile analysis
Pasadena CA, April 24, 2009
Zernike approximation interface
Pasadena CA, April 24, 2009
Comparing profiles
Pasadena CA, April 24, 2009
Interface to create aspherics
Pasadena CA, April 24, 2009
We can show how the tool can be used to estimate power losses in one
reflection due to dust particles. Typical dust pike effect in phase data:
Pasadena CA, April 24, 2009
Profile of disturbance in interferometer phase intensity measurement
Pasadena CA, April 24, 2009
To get power loss estimation let’s remove Zernikes up to 8th order. The rms
decreases to 2.668nm.
Pasadena CA, April 24, 2009
If we consider this Zernike residual as a sample of two-dimensional

centered stationary random process f (r ) with   rms and assume
that HeNe wavelength 632.8nm is much bigger than the Zernike residual
PV = 39.37nm then the Fourier transform
expansion of the reflected wave

from normally impinging beam  (r ) :



2ikf ( r )
 (r )  e
 (r )


can be used just up to the first power of 2ik [ f (r1 )  f (r2 )] . Then we get
scattering losses from the power spectrum density:
  4 k 2 2
that in this case is 0.05.
It would be interesting to get detailed information about field spectral
variations depending on dust pike position and its shape parameters,
though it is doubtful if the Fresnel diffraction or PDE will work at all. More
direct involvement of the Kirchhoff equation looks necessary.
Pasadena CA, April 24, 2009
I would be happy to answer your
questions if you have some.
Thank you.
Pasadena CA, April 24, 2009