Neng-Tien Huang

A New Direction in Computational Electromagnetics: Solving Large Problems Using the Parallel FDTD on the BlueGene/L Supercomputer Providing Teraflop-Level Performance

Rapid developments in high-performance supercomputers, with upward of 65,536 processors and 32 terabytes of memory, have dramatically changed the landscape in computational electromagnetics. The IBM BlueGene/L supercomputer are examples. They have recently made it possible to solve extremely large problems efficiently. For instance, they have reduced 52 days of simulation on a single Pentium 4 processor to only about 10 minutes on 4000 processors in a BlueGene/L supercomputer. In this article, we investigate the performance of a parallel Finite-Difference Time-Domain (FDTD) code on a large BlueGene/L system. We show that the efficiency of the code is excellent, and can reach up to 90%. The code has been used to simulate a number of electrically large problems, including a 100 * 100 patch antenna array, a 144-element dual- polarized Vivaldi array, a 40-element helical antenna array, and an electronic packaging problem. The results presented serve to demonstrate the efficiency of the parallelization of the code on the BlueGene/L system. In addition, we also introduce the development of the high-performance Beowulf clusters for simulation of electrically large problems.

Analysis of multiple FSS screens of unequal periodicity using an efficient cascading technique

In this paper, we present an efficient cascading procedure for analyzing frequency selective surface (FSS) systems consisting of multiple FSS screens of unequal periodicity embedded in multiple dielectric layers. In this procedure, we first find a global period for the FSS system by studying the composite in its entirety. Next, we compute the scattering matrix [S] of each of the FSS subsystems for the global Floquet harmonics by applying a relationship we establish that maps the [S] matrix of the subsystem for the individual Floquet harmonics to that for the global harmonics. This mapping-cum-filling process substantially reduces the effort needed to compute the [S] matrix of a subsystem. Finally, we compute the [S] of the entire system by applying a modified cascading formulation, in which one matrix inversion step is eliminated, resulting in a reduction in the total computing resource requirement as well as time. Two numerical examples are given to illustrate the efficiency and effectiveness of the technique.

Parallel FDTD Modeling of a Focal Plane Array with Vivaldi Elements on the Highly Parallel LOFAR BlueGene/L Supercomputer

This paper describes some preliminary results of a numerical study towards the modeling of a focal plane array (FPA) with Vivaldi elements, carried out on the LOFAR BlueGene/L, a supercomputer with more than 6000 nodes located at the University of Groningen. The parallel finite-difference time domain (PFDTD) code was used to simulate the array because it is able to achieve very high efficiency on such parallel computers

Numerical and experimental studies of a dual-polarized planar connected-array antenna for the Australian Square Kilometer Array Pathfinder

Good agreement with measured radiation patterns of the connected-array antenna have been obtained by use of GEMS, MWS and an efficient CBFM approach that we have developed. The results give confidence in this component of the FPA system-modeling capability.

Numerical study of a dual-polarized Focal Plane Array (FPA) with Vivaldi elements placed in the vicinity of a large feed box using the parallelized FDTD code GEMS

Dense Focal Plane Arrays (FPAs) are currently regarded as one of the strongest candidates for use in next generation radio telescopes, such as the Square Kilometer Array (SKA). Since neither the conventional waveguides nor horn arrays are able to realize the full potential of the SKA, the Vivaldi array, comprising of electrically small wideband tapered slot antennas, has been designed to fill this need. An initial version of the 144 elements dual-polarized Vivaldi element array, whose design is based on the use of the dense FPA concept, has been realized at ASTRON and has been successfully tested on the Westerbork Synthesis Radio Telescope (WSRT) [1], [2]. The tests have demonstrated that it is feasible to build a wideband FPA with illumination efficiency superior to that of a conventional waveguide/horn.

A robust parallel conformal FDTD algorithm and its application to electrically large antenna array simulation

In this paper, we describe an optimized parallel conformal FDTD technique in which the field exchange is carried out between the neighboring processors in a highly efficient way. A robust conformal mesh design technique, based on the proposed parallel scheme, is also developed. In addition, this paper introduces a novel approach to realizing the FDTD excitation and collecting the computed field values in the parallel processing scheme. Several representative examples are presented to validate the proposed techniques.

Analysis of FSS composites comprising of multiple FSS screens of unequal periodicity

In this paper, we have presented an efficient cascading procedure for analyzing an FSS composite system, comprising of multiple FSS screens of unequal periodicities, embedded in multiple dielectric layers. The numerical examples demonstrate the efficiency and effectiveness of the technique. Based on this procedure, a computer program has been developed for the analysis of dissimilar FSS systems.

Investigation of the vivaldi array for Square Kilometer Array (SKA) application using the parallelized FDTD code GEMS on the LOFAR blue gene/L supercomputer

This paper presents the numerical simulation results of a Vivaldi 9x8x2 array. This type of arrays is being designed by ASTRON for the square kilometer array (SKA). The simulation has been carried out by using general electromagnetic solver (GEMS)-a parallelized finite difference time domain (FDTD) solver (W. Yu et al., 2005)-running on the LOFAR BlueGene/L supercomputer located at the University of Groningen. This supercomputer is one of the worlds fastest as well as largest-it has up to 6000 nodes- and the high parallelization efficiency of GEMS (-90%) makes the combination a very powerful tool for solving large complex problems in a time-efficient manner. The array parameters studied include the input impedance of the active element, coupling between the active element and those surrounding it, and the far field pattern of the same element, when the rest of the array elements are terminated with its characteristic impedance. Numerical results generated from the simulations using GEMS are presented in the paper and are compared with the measurement results.

A novel technique for solving EMI/EMC problem between large antennas mounted on complex platforms

The EMI/EMC problem that affects the co-siting of antennas mounted on complex platforms is a very challenging one, especially when the body on which the antennas are mounted is large - as are the antennas themselves - and the separation distance between them is also many wavelengths in dimension. The objective of this paper is to present a novel technique for handling such large problems. Some examples of such problems are shown that display phased arrays and other antennas mounted on ships and airplane structures. One obvious approach to handling a problem with many degrees of freedom (DOFs) is to simulate it on a large parallel machine, e.g., a Beowulf cluster. Through such parallel processing helps expand the range of problem sizes that can be solved directly, many real-world problems still remain beyond our reach, often forcing us to make unreasonable approximations that cannot be fully justified. To address such coupling problems between arrays, we have developed a technique that modifies the parallel FDTD algorithm (the FDTD is chosen because it is a general-purpose code well-suited for analyzing complex and inhomogeneous structures) in a manner that enables the solving of the original problem in several successive steps, circumventing the need to deal with it in a single simulation, which would be prohibitive to carry out when the problem size is very large

Investigation of instability characteristics arising in the FDTD simulation of electrically large antenna arrays

In this paper, we investigate the instability problem that arises in the FDTD simulations of an electrically large problem. When the configuration to be simulated includes many small features such as thin slots, dielectric slabs of small thickness, thin PEC or curved objects, the problem becomes challenging in terms of memory requirement even when parallel FDTD is used in conjunction with a non-uniform mesh. Additionally, a non-uniform mesh in the PML region triggers some instabilities in the solution when a conventional PML boundary condition is used. In this paper, we propose some practical ways to overcome such instability problems, by using an antenna array to illustrate the application of the proposed techniques.