Computer models mimic shipboard antenna performance
An engineer clicks and rolls a computer mouse on its pad, and a web of gray cylinders, knitted together to resemble the image of a warship, twists and rolls on the computer monitor in response.
From above and behind the ship, the view zooms downward to a mast on the superstructure. The engineer plucks a short wire from one spot on the mast and with another click, edits its location. Presto - a new direction-finding antenna design has just found a new test site.
Thirty years ago, a designer at Southwest Research Institute's (SwRI) Signal Exploitation and Geolocation Division hauled full-scale prototype antennas to a test facility located on the Institute's land-locked campus in South Texas. The division had built a full-scale mock-up of a U.S. Navy FRAM destroyer superstructure using an array of metal plates, pipes, and pylons. On its towers, engineers bolted their antenna prototypes - first at one location and then another, all the while measuring and comparing signal responses to determine the optimum installation on a real ship.
That destroyer mock-up long ago sailed off to the scrap heap as SwRI designers opted for a more practical, less expensive system. In its place has arisen the practice of reduced-scale modeling which, though less bulky and expensive, still requires many experimental configurations during the antenna-location optimization process. The sheer number of variables that can affect signal behavior necessitated the use of rules of thumb and experimentation.
Now, computer-based numerical modeling offers a promise of superior technology and provides significant improvement in the ability to accurately predict prototype performance. It also dramatically reduces the requirement for experimental testing. Computational modeling, in effect, has put the old full-scale model on a desktop.
New tool aids long-standing DF program
The transformation has been relatively swift at SwRI's Signal Exploitation and Geolocation Division. Two years ago an internal research project was begun, focusing on the application of computational modeling to electromagnetic behavior as it relates to the calibration of high-frequency shipboard direction finding (DF) antennas.
DF technology is an internationally recognized Institute strength and one of SwRI's longest-lived research programs. Since the 1950s, SwRI has designed, manufactured, installed, and maintained shipborne DF systems for navies of the United States as well as several of its allies. However, regardless of the changing methodologies of design and development, the final measure of success for a new antenna array, whether past or present, is in its actual shipboard performance.
When a newly designed antenna array is ready for installation, it is calibrated on the ship for which it was designed. Readings are taken roughly in a circle around the vessel to measure the system's response to signals at different frequencies and azimuths of arrival. Once that data bank, called an array manifold, has been assembled and digitally recorded, it is entered into the ship's DF system database and used as a reference for interpreting signals during actual operations. The signal comparison uses an Institute-developed beam steer vector match (BSVM) algorithm, which derives a beam form response for the incoming signal and compares that response to signals of known azimuths of arrival stored in the array manifold.
This calibration method is quite effective at capturing the unique electromagnetic nuances of a specific vessel. However, even among ships of the same class, slight differences in assembly and layout may lead to as much as several meters' difference in antenna location, and taking the ship to sea for calibration can cost hundreds of thousands of dollars.
Numerical modeling of the array manifold
SwRI engineers reasoned that if the number of potential antenna locations could be narrowed prior to testing, time and costs would be saved. An internal research project compared array manifold data that was gathered in the traditional, empirical manner to data from a manifold generated via computer modeling. The initial goal was to compare how well the two methods would agree in their ranking of various potential DF array locations on the ship.
Preliminary results from the ongoing research indicated that the computer modeling approach and the traditional approach coincided well in ranking the array possibilities. Moreover, and somewhat unexpectedly, the computer model also proved to be substantially accurate at predicting the actual DF performance of the array. Exactly how accurate, remains a subject of ongoing project evaluation.
Numerical modeling has been completed for three ships with different levels of detail. Each is being evaluated to determine the level of detail required to predict accurately the DF array beam characteristics. Individual numerically simulated antenna responses have compared well in both magnitude and phase to range-measured responses. Also, a statistical model has been developed to predict the probability of large DF errors as a function of the number of antenna elements, the signal-to-noise ratio, and the general characteristics of the ideal DF response. The numerical model permits DF performance to be estimated during array design.
Progress in computer power, modeling software
Today, engineers are benefiting from a trend toward
lower-priced and higher-powered desktop data processors. A few years ago, one needed a
$20,000 to $40,000 workstation to perform such complex analyses. Today, a 200 MHz
processor with 256 MB of memory located in an office environment can perform the required
computational modeling within reasonable standards of time and accuracy. An entire ship
can be analyzed in the high-frequency range (2 to 30 MHz) in about an hour of computer
time. The time requirement increases as higher frequencies are modeled, because more
numerical segments must be created and analyzed.
A modeling segment is based on 0.10 of wavelength, which varies inversely with frequency. For a 30 MHz system, the wavelength is about 10 meters, which requires modeling in segments of one meter. Higher frequencies have proportionally shorter wavelengths, which means that a numerical model of a ship's electromagnetic behavior at those frequencies must be constructed using smaller, more numerous segments.
U.S. Navy-developed software such as the Numerical Electromagnetics Code (NEC) and Structure Interpolation and Gridding (SIG) provide modeling capabilities while visualization programs such as POV-RayTM bring the project to reality by providing three-dimensional visualization. All can be provided to the engineer for a few thousand dollars.
Recent advances in computer graphics make it easier for the operator to "see" a model as the computer sees it. The visualization displays a ship as a network of tubes of varying diameters. Each tube represents a computational analysis of the ship's electromagnetic characteristics when modeled as electric current flowing through wire. The varying diameters of cylinders can mimic the signal reflectivity and electromagnetic response of various components of a ship, and consequently the potential interference with radio signal reception at a given location. As a result, finding an optimal antenna location aboard a compact vessel becomes a matter of relationships. Placing a DF antenna array near a large rotating metal radar antenna or near a metal hatchway that opens and closes frequently can cause significant changes in electromagnetic behavior, which in turn may affect antenna performance. Scale-model testing continues to be conducted to validate both computer modeling and full-scale measurements.
While experimental testing of new antenna array designs remains a valuable resource for shipboard DF, a computer model is valuable as a "first-look" technology for narrowing the list of potential locations in the search for an optimum site. As computers continue to increase in power and engineers are able to exploit computational modeling with more variables, future design models will be likely to resemble less a simplified system operated in a laboratory, and more the complex systems that operate in the real world.
Along with these capabilities will come the ability to allow the computer to automatically iterate through the many possibilities and optimize the design on its own, thereby giving the engineer an even more powerful and efficient tool for realistic design and analysis.
Published in the Summer 1998 issue of Technology Today®, published by Southwest Research Institute. For more information, contact Joe Fohn.