One Day Workshop on "Optimization Engines for Wireless and Microwave Computer Aided Engineering"

June 20, 2002, Carleton University, Minto CASE Bldg Room 2014 , 08:30-18:00

      Refreshments and lunch will be provided.

Please respond to j.bandler@ieee.org by June 11 if you will attend.

WORKSHOP SCOPE

We present new results in optimization technology for RF, wireless and microwave circuit design, integrating electromagnetic (EM) simulations.  Atutorial review, new research directions, algorithms and software engines are presented.  Applications exploit commercial EM simulators such as Sonnet, Agilent ADS and HFSS: waveguide structures, microstrip filters, patch antennas, circuit decomposition techniques, mixed EM/circuit structures and signal integrity optimization of VLSI packages and interconnects.

PART I: SPACE MAPPING

Space mapping optimization intelligently links companion "coarse" (ideal or low-fidelity) and "fine" (practical or high-fidelity) models of different complexities.  Examples include full-wave electromagnetic simulations with empirical circuit-theory based simulations, or an engineering device under test coupled with a suitable simulation surrogate.  The methodology follows the traditional experience of engineers yet appears to be amenable to rigorous mathematical treatment.The exploitation of managed "space mapped" surrogates promises significant efficiency in engineering design. The original concept of space mapping, and the aggressive space mapping approach to engineering design optimization will be discussed. Our latest trust region aggressive space mapping optimization algorithms will be discussed.  Artificial neural network (ANN) approaches vs. generalized space mapping for device modeling from EM data will be mentioned.  We discuss "neuro space mapping" approaches to device modeling and circuit optimization using ANN methodology.  An expanded space mapping design framework exploiting preassigned parameters will be discussed.  We review the new Implicit Space Mapping (ISM) concept in which we allow preassigned parameters, not used in optimization, to change in some components of the coarse model. We have devised novel physical examples to illustrate our ideas: Cheese Cutting Problem, Shoe Selection Problem, Wedge Cutting Problem, etc.

PART II: NEURAL MODELING

Neural networks have gained attention in RF/microwave CAD.  Trained neural network models can be used in place of detailed physics/EM models to speed-up microwave design.  An automatic model generation technique will be presented exploiting the neural network approach. For user-required model accuracy, the algorithm automatically drives data generators to train neural network to learn the required model behavior.  The technique automatically determines the number of samples and their distribution in model input-space using an adaptive sampling algorithm.  Dynamic neural network methods will be presented for large signal modeling of nonlinear devices and circuits.  The technique allows the model to learn the nonlinear behavior of device/circuit input-output relationships through time-domain waveform and/or frequency domain spectrum data.  The trained nonlinear model can be incorporated into high-level simulation, providing speed and efficiency for circuit and systems design.

PART III: ADJOINT SENSITIVITIES FOR EM SIMULATIONS

The adjoint variable method (AVM) for design sensitivity analysis (DSA) offers superior computational efficiency and accuracy in comparison with other techniques for sensitivity estimation.  However, its implementation with full wave EM solvers is not trivial and its accuracy depends on a number of factors.  We show how this method is applicable in conjunction with any EM computational method that reduces the problem to a system of linear equations in the frequency domain.  It is shown that the solution to the adjoint problem can be obtained with little overhead once the original problem is solved.  We derive the general design sensitivity formula, which produces the gradient of the objective function through a single analysis regardless of the number of the design parameters. The concept is illustrated by applications based on the Method of Moments (MoM).  We discuss factors that influence accuracy.

PART IV: NeuroModeler DEMONSTRATION

The world's first software for neural based device modeling and circuit design.  Examples include EM simulations, filters, transmission-line networks, transistor device models, etc. NeuroADS and advanced neural network based tools for RF/microwave modeling, automatic model generation for RF/microwave devices, neural models for passive and active components, and their use in circuits and systems design.

 

Workshop content will include:

IEEE MTT-S INTERNATIONAL MICROWAVE SYMPOSIUM SEATTLE, WA, JUNE 2002

TECHNICAL PRESENTATION HIGHLIGHTS

Papers:

J.W. Bandler, Q.S. Cheng, N. Georgieva and M.A. Ismail, "Implicit space mapping EM-based modeling and design using preassigned parameters," Session WE1E, June 5.

N. Georgieva, S. Glavic, M. Bakr and J.W. Bandler, "Feasible adjoint sensitivity technique (FAST) for EM design optimization," Session WE3D, June 5.

J. Xu, M.C.E. Yagoub, R. Ding and Q.J. Zhang, "Neural based dynamic modeling of nonlinear microwave circuits," Session WE4D, June 5.

V. Devabhaktuni, B. Chattaraj, M.C.E. Yagoub and Q.J. Zhang, "Advanced microwave modeling framework exploiting automatic model generation, knowledge neural networks and space mapping," Session WE4D, June 5.

J.W. Bandler, A.S. Mohamed, M.H. Bakr, K. Madsen and J. Sondergaard, "EM-based optimization exploiting partial space mapping and exact sensitivities," Session IF-TH-30, June 6.

Workshop Organization:

J.W. Bandler, Organizer, "Microwave Component Design Using Space MappingMethodologies, " Workshop WMB, Monday, June 3.

Workshop Presentations, WMB, June 3:

J.W. Bandler and Q.S. Cheng, "Space mapping approaches to EM-based device modeling and component design."

J.E. Rayas-Sanchez and J.W. Bandler, "EM-based design through neural space mapping methods."

M.H. Bakr, J.W. Bandler, K. Madsen and J. Sondergaard, "Theory and applications of the space mapping technique."

Q.J. Zhang, V.K. Devabhaktuni, J.J. Xu and M. Yagoub, "Knowledge based neural network approaches for microwave modeling and design."

Exhibition, June 4-6, 2002:
Q.J. Zhang, Exhibition Booth 1026, NeuroModeler, NeuroADS and advanced neural network based tools for RF/microwave modeling, automatic model generation for RF/microwave devices, neural models for passive and active components, and their use in circuits and systems design.

Microwave Applications Seminar:
Q.J. Zhang, Neural Network Aided Microwave modeling and design.

 

 

2002 WORKSHOP PROGRAM SCHEDULE

 

08:30               Coffee

 

08:45               J.W. Bandler (McMaster and President, Bandler Corporation "Opening remarks”

 

08:50               Dr. Prakash Bhartia (Director General, DREO), “Welcome address”

SPACE MAPPING OPTIMIZATION TECHNIQUES

09:00               J.W. Bandler (McMaster and Bandler Corporation), “New developments in optimization technology for RF, wireless and microwave circuit design, integrating electromagnetic (EM) simulations”

10:00               A.S. Mohamed (McMaster), “Space mapping optimization exploiting exact sensitivities”

10:30               Coffee break

10:50               Q.S. Cheng (McMaster), “Theory and applications of implicit space mapping using preassigned parameters”

11:20               S. Dakroury (McMaster), “Trust regions and aggressive space mapping: new illustrations and insight”

GUEST SPEAKER

11:40               Dr. Jim Rautio (President, Sonnet Software), “Adaptive band synthesis—robust extraction of wide-band high resolution data from electromagnetic analysis”

12:30               Lunch

NEURAL MODELING

13:30               Q.J. Zhang (Carleton), “Neural networks for high-frequency modeling and design, an overview”

14:10               V.K. Devabhaktuni (Carleton), “Automatic generation of RF/Microwave models”

14:40               J.J. Xu (Carleton), “Dynamic neural networks for modeling of nonlinear devices”

15:10               Coffee

ADJOINT SENSITIVITIES FOR EM SIMULATIONS

15:30               N.K. Georgieva (McMaster), “Feasible adjoint sensitivities for efficient gradient-based optimization with frequency-domain EM solvers”

16:10               G. Shen (McMaster), “Challenges in the sensitivity analysis with time-domain EM solvers"

 

16:40               NeuroModeler DEMONSTRATION

 

17:20               J.W. Bandler (McMaster and Bandler Corporation), Closing Presentation and Overview of the Day, “The bright future of space mapping: a sensible approach to engineering optimization”

 

18:00              End

 

Guest Speaker: Dr. Jim Rautio, Sonnet Software, "Adaptive Band Synthesis -- Robust Extraction of Wide-Band High Resolution
Data From Electromagnetic Analysis"

 

Abstract

The interpolation of electromagnetic data has been available in the industry for some time.  By using results from analysis at just a few frequencies, a full frequency response can be obtained using these techniques.  The limitations in present approaches have limited their acceptance into the mainstream design process.  These limitations include poor dynamic range, limited bandwidth especially for complex high Q structures, and frequent failure to converge.  These limitations can be almost entirely overcome by extracting certain critical information from the moment matrix used internally by the Sonnet electromagnetic analysis.  This information is then used to build a more sophisticated model. For example, an entire filter response with 120 dB dynamic range and a complicated pass-band response can be obtained from analysis at under a half dozen frequencies.  A general overview of this approach will be given along with example results.  SonnetLite, the free version of Sonnet, includes this capability, CD's will be available at the presentation.

Biography of James C. Rautio

James C. Rautio received a BSEE from Cornell in 1978, a MS Systems Engineering from University of Pennsylvania in 1982, and a Ph. D. in electrical engineering from Syracuse University in 1986. From 1978 to 1986, he worked for General Electric, first at the Valley Forge Space Division, then at the Syracuse Electronics Laboratory. At this time he developed microwave design and measurement software, and designed microwave circuits on Alumina and on GaAs. From 1986 to 1988, he was a visiting professor at Syracuse University and at Cornell. In 1988 he went full time with Sonnet Software, a company he had founded in 1983. In 1995, Sonnet was listed on the Inc. 500 list of the fastest growing privately held US companies, the first microwave software company ever to be so listed. Today, Sonnet is the leading vendor of 3-D planar high frequency electromagnetic analysis software.  Dr. Rautio was elected a fellow of the IEEE in 2000 and received the IEEE MTT Microwave Application Award in 2001.

For updated information please visit http://www.sos.mcmaster.ca/upcomingevents.htm