Before we talk about specific solvers, we would like to take a moment to go through a bit of our solver development approach. New hardware designs are constantly pressured for better performance, lower cost, and less development time. So, to be helpful and relevant in this increasingly demanding environment we are producing, or in many cases co-producing solvers with some of the best solver minds in the world. STAAR is constantly working to bring the most sophisticated solver technology to commercial markets with features such as: advanced iterative capabilities, parallel processing, and advanced numerical techniques.
What does this mean for you as a user of these solvers? Well, improved iterative solution technology means faster, more rapid convergence which saves computational resources and gets you answers faster. Our developing support for parallel processing allows you to solve bigger more complex problems using multi processor PCs, UNIX machines, all the way up to massively parallel Cray or IBM systems. This computing power can now readily be harnessed on problems that were impractical to try solving before. Advanced numerical techniques in statics, eigenmode, frequency-domain and time-domain solvers yield answers more quickly and more accurately than before. All of which will translate into better designs, lower costs, and shorter design cycles. Finally, we wrap up all of our components with a high level of integration to make access to these state-of-the-art capabilities much more intuitive and user friendly than you will see from anyone else.
Eigenmode
With the parallelized eigensolver Omega3p developed by the Numerical Methods Group at Stanford Linear Accelerator Center, several numerical methods are combined in an innovative way to permit the solution of extremely large and complex problems. The approach yields far superior accuracy, convergence, and scalability as compared to other algorithms. It is wrapped with a sophisticated, user friendly interface to offer truly exceptional accelerator component design capabilities. Commercially available since Analyst Version 1 was released in August, 2000. A 2-D eigensolver (Omega2) has also been developed by SLAC and is available within Analyst.
To address a wider range of eigenmode problems we have developed a 3-D parallel eigensolver called OM3P that has both standard Lanczos for smaller problems and the Omega3P hybrid iterative method for large problems. OM3P was introduced in Version 5, and it includes sophisticated techniques for extracting closely spaced modes in mode clusters, and degenerate modes, and has specialized finite-element basis functions that dramatically reduce the computational time required to solve very large eigenproblems.
Driven Frequency
With Version 4 of Analyst we introduced RF3P, a new parallel driven frequency electromagnetics solver. RF3P allows arbitrary numbers of ports and modes per port, and supports dielectric, permeable, and lossy media. RF3P will calculate the generalized scattering matrix for a structure, and allows frequency stepping to create response curves. Fields may also be output for both port modes and within the volume.
Electrostatics
There are two electrostatics solvers in Analyst - ES3P and Extractor. Extractor has been available in Analyst since Version 1, and it is a hybrid finite-element/boundary-element code that is applicable to "open" problems in which fields and the capacitance matrix are desired on external surfaces of conductors. ES3P was introduced in Version 5, and it is a parallel, finite element-based electrostatics solver that is useful on "closed" problems, that is, where the boundaries can be modeled using either Neumann or Dirichlet conditions. ES3P computes volumetric potentials, electric fields, and charge distributions on conductors. Both Extractor and ES3P can model dielectric media.
Time Domain
Time-domain analysis is useful in efficiently getting spread-spectrum information about devices. To get highly accurate information from an analysis, conformal meshing at a very fine level must be used. This, in turn, makes the element count enormous which results in the only practical solution of applying a parallelized code to the problem if one would like a convergent solution in a reasonable time. With the parallelized Tau3p solver, developed by the Numerical Methods Group at Stanford Linear Accelerator Center, sophisticated, robust numerical methods are introduced to permit the solution of these extremely large and complex problems. Superior accuracy, convergence, and scalability as compared to other algorithms are hallmarks. It is wrapped with a sophisticated, user friendly interface to offer exceptional capability to analyze and optimize beamline components. Commercially available in the 3rd quarter of 2003.