Essais & Simulations n°108

Compatibilité

Compatibilité électromagnétique Cages anéchoïdes Efficient Simulation of Anechoic Chamber Designing anechoic chambers is very expensive and time consuming in case the design only relies on physical prototype testing. This is the reason why companies that build semi-anechoic chambers have started to consider the virtual prototyping as a good candidate to save cost and to improve the efficiency of their designs. But modeling of anechoic chambers is a very difficult task as it implies Electromagnetic simulation of a large volume at relatively high frequencies and including materials with very high permittivities and permeabilities. Simulation is critical in designing these chambers because nearfield effects in the 30 to 200 MHz range cannot be determined by theoretical methods. A very fine mesh is normally required in the wall area to model the performance of absorbers that are used to make the chamber act as if it were free space. The fineness of the mesh typically results in very long simulation times, such as the 15 weeks that could be needed on a desktop computer in the past (before 2004) to model chambers to predict the performances. Gwenal Dun, RD Engineer for Siepel (Ph. D.), used a variety of different electromagnetic simulation tools to address this challenge in the past but ran into problems with both poor accuracy and long Figure 1 : Antenna test setup compute times. We then worked together the developers of CST MICROSTRIPES electromagnetic simulation software, to implement a feature that makes it possible to model the ferrite absorbers used in the chamber as a boundary condition rather than part of the computational domain. This change made it possible to increase mesh size by a minimum factor of 15, reducing compute time by more than 95%. The simulation results provided a nearperfect match to physical testing. Development of semi-anechoic chambers International regulatory agencies have greatly increased radio frequency (RF) emissions and susceptibility requirements since they were first introduced in the 1970s. Generally the standards on RF emissions are based on tests performed outside on an OATS but, these suffer from the effects of weather conditions and ambient noise. To overcome the problem of weather conditions and ambient noise, semianechoic chambers have been developed as shown in Figure 1. The chamber is a RF shielded box with the walls and ceiling lined with materials that are highly absorbent of RF waves in order to provide conditions similar to an OATS. Today, regulatory agencies allow most products to be tested for EMC in semianechoic chambers rather than OATS. They require, however, that these chambers behave in a way that closely corresponds to OATS. The American ANSI C63-4 and the European EN50147-2 standards require that EMC testing be performed in a chamber where the Normalized Site Attenuation (NSA) deviates from an OATS by no more than ±4 dB. The design challenge Companies that build semi-anechoic chambers must be certain that their products meet this specification. Physical testing provides a poor solution because it is very expensive to build a prototype chamber and the physical testing required to evaluate the performance of the chamber over the full range of required frequencies and in all areas of the chamber would cost too much and take too long. Theoretical approaches provide good results for certain subsets of the problem but do not work for others. For example, at very high frequencies, typically above 1 GHz, the antenna geometry is not important so the electromagnetic field can be calculated based on the antenna radiation pattern and on the reflectivity of the wall. But this approximation does not apply to lower frequencies, where E S S A I S & S I M U L AT I O N S ● O C TO B R E , N OVEMBRE, D É C E M B R E 2 0 1 1 ● PAG E 2 3

Compatibilité électromagnétique the geometry of the antenna is very important due to the near field effect and simulation is a must. It appeared that improving the simulation process was critical to optimizing the performance of chambers so Siepel decided to carefully evaluate the leading electromagnetic simulation methods in terms of their ability in this area. Frequency methods such as Method of Moments (MoM) do a good job of simulating the wire antennas used for the qualification of anechoic chambers but cannot accurately simulate the walls of the chamber due to very high memory and CPU time requirements. On the other hand, finite difference time domain (FDTD) methods work well for the walls but have difficulty in modeling wire antennas, which typically require a mesh of 1 mm or less. Models with meshes this small typically have solution times measured in months, which is far too long to have a positive impact on the design process. TLM method provides accuracy and speed Figure 2 : CST MICROSTRIPES model Finally, Mr. Dun turned to the CST MICROSTRIPES implementation of the transmission line method (TLM) from CST for solving Maxwell’s equations Which is now part of the transient solver of CST MICROWAVE STUDIO (CST MWS). The TLM method solves for all frequencies of interest in a single calculation and therefore captures the full broadband response of the system in one simulation cycle. A further advantage is that the TLM method creates a matrix of equivalent transmission lines and solves for voltage and current on these lines directly. This uses less memory and CPU time than solving for E and H fields on a conventional computational grid. The solver tolerates rapid changes in grid density, large aspect ratios of grid cells and localized gridding, enabling the mesh requirements to be kept to an absolute minimum. Finally, an intuitive easy-to-use graphical user interface, optimized meshing algorithm and parallel processing for increased speed, make the software suitable for solving extremely complex and electrically large problems. CST MICROSTRIPES provided the best mix of accuracy and computational efficiency for modeling EMC chambers with ferrite absorbers. We found that the TLM method successfully modeled both the antennas and the chamber itself. We were able to create compact models of antenna structures that reduce the size of the resulting model while maintaining high levels of accuracy. We defined the transmission parameters by the scattering parameters of the balun and the simulation results of the wires. Because baluns can’t be modeled easily S-parameters were used which do not influence electromagnetic propagation. The use of a compact model to represent the antenna meant that the smallest element size required was 15 mm for the wire connection. Special boundary condition overcomes problem But we had to overcome the problem in modeling the walls of the chamber. The ferrite absorbers SIEPEL FE30Z used in the chamber are only 6.7 mm thick, which meant that a mesh of 1 mm was needed. Reducing the mesh size to this level would require a 15 week simulation time. This was much too high so we investigated whether there was a way around the problem. We worked to develop a special boundary condition that simulates the reflectivity of the ferrite absorbers, eliminating the need to include them in the model. The boundary condition was defined by the frequency dependent surface impedance of a one dimensional TLM ladder network and defined at the air-ferrite interface for the two polarizations of the E field parallel and perpendicular to the to the air/ferrite interface. This limit condition takes into account the incidence angle and the polarization of the electromagnetic wave. The key advantage of making the walls into boundary conditions is the elimination of the need for the 1 mm mesh in this area. This means that the most critical area is the antenna connection which only requires a 15 mm mesh. The resulting increase in the mesh size reduced the computation time to only 1 week on a desktop computer, which was fast enough to serve as the primary evaluation tool during the design process. The boundary condition had no effect on the accuracy of the simulation. To validate the model, simulation and measurement results were compared for the two polarizations and two heights of the emission antenna. The E S S A I S & S I M U L AT I O N S ● O C TO B R E , N OVEMBRE, D É C E M B R E 2 0 1 1 ● PAG E 2 4