Complex Antenna System Design
By Jim DeLap, Dane Thompson, Ph.D., and Nick Hirth, Ansoft Corporation

From one to many: Complex radiating elements like planar inverted F antennas (PIFA), patches, flared notches, ultrawideband radiators, electronic bandgap elements, and emerging metamaterial antennas are often designed using powerful 3-D electromagnetic field solvers. Traditionally, only single elements or small collections of elements have been simulated using finite element or method-of-moments field solvers. New developments in field solver technology enable much larger simulations and more comprehensive design and optimization, which include not only the antenna but also the platform or environment within which the antenna operates.

In this article, three key enabling developments are presented that make it practical to simulate complex antenna systems. Case studies of a phased array antenna with feed network system and a slot antenna with radome system will be used to illustrate the technologies.

Development 1: Combination of Circuit and 3-D Electromagnetic Simulations
Perhaps the most significant development in field solver technologies is the creation of a two-way link between traditional circuit simulation tools and 3-D electromagnetic simulation tools. Previously, designers would design and simulate the antenna and any supporting circuitry independently. For example, designers would develop an antenna’s feed network in a circuit simulation tool, optimize the design in simulation software, fabricate a prototype, and then measure the actual performance of the circuit. Similarly, when considering the antenna portion of the system, a designer would construct a single element in a 3-D simulation software package, enter a set of idealized inputs, and simulate some performance characteristic such as far-field radiation or input impedance. Eventually, the feed network and the antenna would be brought together as physical hardware and the system would then be tested and measured. The weakness of this methodology is that the designer does not get to observe the actual interaction between the feed network and the antenna until the physical hardware has been created. This is a less than ideal time for modifications to be made. Changes at this point can be costly and time consuming. Fortunately, the situation is changing. Now, designers have the ability to simulate both the feed network and the antenna as a single system. System optimization can now occur at the simulation stage. Let’s look at a specific example.

In Figure 1, a fairly typical corporate feed network for a phased array is presented. Working from the left, one finds for each phased array element a series combination of isolators, LNAs, and phase shifters. Each of these blocks is built in an analog circuit simulation environment. The output of this series sub-circuit is fed into a power combiner. In a given column, two adjacent cells are combined in a power combiner (point (A) in Figure 1). Individual pairs are combined (B) until all cells in a column are connected. Columns are then connected by another set of combiners (C). Normally, this circuit would be simulated, built, and then tested independently of the antenna shown in Figure 2. Now, 3-D structures can be added and their interaction with the circuit elements can be simulated. For example, the isolators (D) were first created and simulated in a 3-D electromagnetic solver and then inserted into the circuit simulator. Similarly, the interconnects (blue lines) could have been created in the same way. This level of model integration provides accuracies and predictability that were not previously achievable.

The two-way channel can work in the reverse direction. That is, simulation results from the circuit model can be used to drive the 3-D electromagnetic model. Assume, for example, that the feed network shown in Figure 1 is not in “receive” mode, but rather in “transmit” mode. In transmit mode, the circuit topology in Figure 1 would be flipped 180 degrees. In this instance, energy would flow from right to left in Figure 1 and energy would radiate out of the antenna. Because the antenna and feed network models are connected, the energy applied to each individual element can be adjusted. By adjusting the amplitude and phase applied to each element of the array, the whole system can be optimized for a specific objective. In our case, the design objective was to achieve a beam with all sidelobe gains less than -30 dB. Through experimentation, it was discovered that this objective could be achieved with a specific amplitude and phase combination fed or “pushed” into the (circular waveguide) antenna elements as shown in Figure 3.

In Figure 4, the gain results before and after the optimized amplitude and phase combination were fed into the antenna have been plotted. The implications of this system-level simulation are far-reaching. Now, designers have a tool with the requisite sophistication to test system-level interactions and performance. By adjusting energy inputs in the circuit model or changing physical dimensions in the 3-D model, for example, system-level trade-offs can be dynamically simulated and tested. Moreover, adjustments can be made parametrically. With this capability, optimal designs can be identified in simulation – which may inspire huge cost savings or allow additional design margin or performance.

Development 2: Solving Large Simulations by Dividing into a Series of Smaller Projects
One of the common challenges with CAD-based electromagnetic simulations is the memory and time required to obtain solutions to large problems. In many instances, the on-board RAM is insufficient to allow for a convergent solution in a reasonable time. It is not uncommon for a complex simulation to take hours to solve on a single PC. One way to avoid these problems is to use a divide and conquer approach, whereby the design is divided into smaller pieces and the simulator (field solver) is used to solve the smaller pieces individually. The project is then reassembled by linking the component projects in daisy-chain fashion. The output of the first project becomes the input to the second project and so on. The advantage of this approach is that the total RAM and solution time for the individual pieces is less than the time it would take to solve the system if the project were considered in toto. By reducing the total system into bite-sized pieces, designers reduce the volume that finite element field solvers have to mesh and, in turn, reduce the memory and time required to obtain a convergent solution.

The basic concept is illustrated in Figure 5. In this example, the system is a combination of the near field effects of a radiating horn antenna coupled with the far field effects in vacuum. On the left is a horn antenna radiating at some frequency across three cylindrical metal scatterers. The fields shown are a horizontal slice through the center plane of the horn antenna. This “source project” was solved separately and the solution for its right side boundary becomes the input to the target project’s left side boundary. With this input, the target project is then solved separately. In this case, the target project is free space and the fields shown are in the same horizontal plane as the source project. The compute time for the individual solutions is less than the time it would take if both pieces (“volumes”) were simulated in a single project.

Let’s consider another more complex case frequently encountered in antenna design. In this example, the goal is to simulate the interaction of a slot antenna with a conical radome. In Figure 6, the antenna-radome system being simulated is shown. Through symmetry, the mesh volume was further reduced by considering only ¼ of the radome and applying appropriate boundary conditions as shown in Figure 7. The slot antenna is the source to the radome target. That is, the slot antenna’s boundary field solution becomes the “forced” input boundary conditions to the radome. This is illustrated in Figure 7. Each project is solved separately. In Figure 8, the solved fields for the radome interior and the total array far-field with radome effects are presented.

Development 3: Reduction in Simulation Times by Distributing Solutions across a Compute Cluster
Another “divide and conquer” time-saving development is the ability to distribute a single simulation across multiple PCs. A host computer manages the simulation and sends processing requests to client computers. In this manner, multiple geometry parameterizations, optimizations, or frequency sweeps can be processed simultaneously. Returning to the phased array example in Figure 2, we will illustrate this “distributed solve” technique by simulating the frequency response of the antenna with one PC and then again with a cluster of six PCs. In a discrete frequency sweep, the field solver identifies all antenna parameters, for example S parameters, gains, and far field radiation patterns, for a specified set of frequencies. In the test case, six frequencies were selected. The single PC was able to compute the frequency sweep in 3 hours and 11 minutes. The six PC cluster computed the same frequency sweep in 39 minutes -- a 4.9 times reduction in simulation time. Similar tests have shown that simulation time reduction tends to scale linearly with the number of processors in the cluster. By taking advantage of an available compute cluster in this fashion, many projects’ total compute time can be greatly reduced.

Conclusions
Long simulation times and a lack of integration between circuit models and 3-D antenna models have been barriers to complex antenna system simulations. Three key developments in advanced simulation technologies are helping to make complex antenna system simulations possible. First, circuit and 3-D electromagnetic simulations are becoming directly linked. As a result, circuit and antenna systems can be dynamically simulated and tested. In addition, parametric adjustments of the models can be used to test design trade-offs and to identify optimal designs. In our example, we showed how a feed network and a phased array antenna can be simulated together to achieve a system goal of -30 dB sidelobes. Second, advances in our ability to link models are allowing finite element field solvers to reduce large project simulation times and use less memory. By dividing a large, complex project into a series of smaller, linked steps, designers reduce the volume that finite element field solvers have to mesh in any given step and therefore, reduce the project’s overall memory requirements and simulation time. In our example, we showed how memory and simulation time of a slot antenna and radome system could be reduced by first simulating the slot antenna in isolation and then using these results as inputs to a subsequent simulation of the radome. Third, the ability to distribute a simulation project across multiple processors and across a company’s compute farm is also reducing large project simulation times. In our example, we showed how the simulation time of a frequency sweep of a phased array antenna can be reduced by almost 5 times by switching from a single PC to a small cluster of six PCs.

ANSOFT
www.ansoft.com
TXTLINX.COM105
Email this article to a friend!