Recently Antenna Products Corporation faced the challenge of designing an antenna for a military application in the 5 to 6 GHz region. The application required an omnidirectional azimuth pattern and a relatively narrow beam in elevation. The antenna would have been difficult and expensive to design using conventional build and test methods. Joe Ippolito, Engineering Manager for Antenna Products Corporation, used the Transmission Line Matrix (TLM) method of electromagnetic (EM) simulation to model a range of antenna designs. Ippolito identified the best three alternatives and presented them to the customer to make the final choice. “The customer selected one design and when it was built the results matched the simulation nearly perfectly, meeting all of the requirements for the application,” Ippolito said.

Antenna Products Corporation designs antennas that cover the RF spectrum from the kilohertz to 6 Gigahertz range and support countless mission critical military, air / sea navigation, ILS and voice / data communications applications. Antenna Products Corporation also designs and manufactures a product line of antenna accessories including towers, telescoping masts, rotators, rotator controls, HF baluns and tower safety climbing equipment. Phazar Antenna Corporation is a division of Antenna Products Corporation and manufactures a complete line of commercial wireless antennas for the telecom and wireless data industries.

Design challenge

The high directivity requirement of this antenna necessitated the use of multiple radiating elements arranged as a linear array. To feed these elements with the phase and amplitude distribution necessary for a good antenna pattern, a corporate divider/combiner network was employed.

This network splits the input power signal via an arrangement of uneven two-way power dividers so that each element receives a precise percentage of the total input power. Between these divider junctions are transmission lines of various lengths to produce the proper time delay in order for each element to be phased properly with respect to its neighbors. A traditional coaxial feed network would have been too large to be practical for this application. Instead, Antenna Products Corporation engineers decided to use printed circuit boards (PCBs) to provide the elements and circuitry for the feeder network.

A simple and economical design for a collinear dipole array is to position the radiating elements in a series. But series-fed arrays have good pattern performance only over a narrow bandwidth because the electrical delay to each element is in phase only at a single frequency. Because of the relatively wide bandwidth requirement of this application, it was determined that a corporate-fed power distribution network would yield optimum results. The corporate divider network is so named because it resembles the organization chart of a corporation.

The corporate divider network arranges dividers/combiners into a hierarchical tree structure. An advantage of this approach is that the dividing structure can be combined with active components on the same substrate to reduce manufacturing costs. The application’s high frequency had the advantage of reducing the size of the PCBs which in turn reduced fabrication costs. In a corporate network the delay lines to each element are distinct and for a uniform phase distribution there is no bandwidth limitation. This results in a more controlled and coherent array factor as well as better VSWR performance.

Generating the initial design

Ippolito used several routines that he wrote in Mathcad based on engineering formulas to estimate antenna geometry and spacing. He used the results to roughly locate the elements and determine the phase and amplitude distribution for each element. In the distant past Ippolito would have built and tested a prototype of the antenna first with only a rudimentary computer simulation as a general guide. The problem with this approach was that the initial design usually did not meet the requirements and never provided optimal performance. Also, the results of physical testing provided relatively little diagnostic information that could be used to improve the design because of limits on the physical measurements that could be taken. So a long and expensive process was needed to develop the finished design.

In the more recent past, Ippolito used frequency methods such as Method of Moments which did a good job of simulating wire antennas but did not account for the effects of dielectric materials such as the PCB substrates and the radome. Nearly every antenna produced by Antenna Products Corporation is covered by a radome on the order of 1/16 to 1/4 inch thickness. At low frequencies the radome has only a small effect on antenna performance but as the frequency rises it plays an increasingly important role.

Ippolito avoided these problems by using MicroStripes electromagnetic analysis software from Flomerics for microwave and antenna design. MicroStripes uses the TLM method for solving Maxwell's equations. A key advantage of the TLM method when applied to antenna design is that it solves for all frequencies of interest in a single calculation and therefore captures the full broadband response of the system in one simulation cycle.

“We selected MicroStripes because it is very easy to use and because its results closely match physical experiments,” Ippolito said. “Before we bought the software we modeled a number of different antennas that we had designed and built in the past. In every case, MicroStripes accurately simulated the performance of the antenna.”

Modeling the antenna

Model of the antenna

Ippolito modeled the initial antenna design in MicroStripes using the simulation program’s ACIS-based modeler to construct the geometry from primitive shapes. Then the software automatically generated the mesh, snapped it to the geometry, and refined it in curved areas and dielectric regions. The software’s multi-grid meshing capabilities enabled fine cells to be localized to the space occupied by the antenna while the surrounding free space region was modeled using a coarser mesh.

The basic model consisted of four PCBs arranged around a central mast at 90° intervals to provide an omnidirectional azimuth pattern. Each board consisted of two dipole antennas and an associated microstrip power distribution network. From the standpoint of the elevation pattern, this component is simply a two-element array and will be referred to as a module. By stacking these modules, greater directivity and narrower beamwidths can be achieved.

In the actual antenna each board was connected via a coaxial cable to a 4-way 0° power divider network. This component was located near the base of the antenna and tied directly to the antenna input connector. It was not included in the model as it had no bearing on antenna patterns but its loss along with the associated cables was included in antenna gain estimates.

Model of PCB structure

Each PCB comprised a circuit and a ground layer. The circuit layer incorporated the microstrip feed network and the dipole matching circuit. The ground layer acted as a return for the microstrip lines and formed the actual dipole and balun.

Simulation results provided design insights

The simulation injected a broadband pulse into the antenna feed and the time signature was captured by stepping through time. Fourier-transformation of this response yielded frequency-domain results across the entire band of the antenna. The model took about 20 minutes to run on a typical Windows PC.

Surface current density as predicted by simulation

The surface current density simulation results provided a graphical representation of the surface currents at a particular point in time on the structure when all boards were excited in phase. The regions of highest current density were depicted in red while the blue regions were relatively devoid of current.

Far-field radiation pattern

The simulation also yielded the far-field radiation pattern. The predominantly vertically polarized pattern was omnidirectional in the azimuth plane and exhibited the pattern characteristics of a two-element array in the elevation plane. The vertical 3 dB beamwidth was on the order of 35°. The directivity of the pattern was 5.5 dBi, but when losses were taken into account the actual power gain dropped to about 4.5 dBi.

Far-field radiation pattern

Summary of directivity results and calculated power gain as a function of frequency in GHz. The vertical axis shows Gain in dBi.

Iterating to an optimized design

Antenna Products engineers then ran a series of simulation iterations in which various numbers of two-element arrays or modules were stacked. The directivity results of these simulations were graphed versus the number of modules used. In actuality, an antenna of any given length would be a complete unit, not an assembly of separate modules. But the modular concept provides a convenient analytical tool. While directivity provides an upper limit, power gain estimates the actual antenna performance by including losses such as microstrip and coaxial line loss, VSWR, polarization and radome loss.

The engineers took the simulation results for a two-module array and imported it into a general array factor program that he wrote to evaluate the effects of varying design parameters such as dipole length. This program multiplied the individual pattern of each module by the array factor to generate the total resulting pattern faster than could be accomplished with a full EM solution. The general array factor program calculated the return loss, antenna gain and azimuth and elevation patterns for a wide range of antenna designs. Ippolito picked the three best designs and presented them to the customer and let them select the one that best met their needs. “The antenna has been built and is in use and the customer loves it,” Ippolito concluded.

For more information, visit the MicroStripes Web site at www.microstripes.com or contact Flomerics Inc.’s U.S. Headquarters at 4 Mount Royal Ave., Suite 450, Marlborough, MA 01752. Ph: 508-357-2012, Fax: 508-357-2013, E-mail: info@flomerics.com, Web site: www.flomerics.com