Figure 1. AMOS 2½D coaxial cell mesh, TSAR 3D cubical cell mesh, and DSI3D unstructured cell mesh show the differences between the code's geometries.
Scott D. Nelson, George Kamin, Roger Van Maren,
Brian Poole*,
Charles Moeller and David Phelps§
*Lawrence Livermore National Laboratory and §General Atomics
Abstract. The combline antenna for plasma heating, as proposed by General Atomics(1), has unique potential for solving many plasma drive problems. The benefit of the combline design is the utilization of the coupling between elements that avoids a more cumbersome multidrive system.
This design is being investigated using computational EM modeling codes in the 100-400 MHz band using resources at General Atomics and LLNL. Preliminary experimental results, using a combline mockup, agree well with 3D modeling efforts including resonant frequency alignment and amplitudes. These efforts have been expanded into an endeavor to optimize the combline design using both time and frequency domain codes. This analysis will include plasma coupling but to date has been limited to antenna effects.
The combline antenna system is modeled in 3D using a combination of computational tools in the time domain, for temporal feature isolation purposes, and in the frequency domain, for resonant structure analysis. Both time and frequency domain modeling details include the Faraday shield elements, the strap elements, and the feed structure.
INTRODUCTION
The Combline Antenna system was analyzed initially to determine the electromagnetic (EM) modelability for use in future design efforts. This included a set of base comparison experimental measurements in LLNL's EM facilities to compare with modeled results for a mockup antenna fabricated at LLNL. As part of this effort, several modeling codes were used in the verification activities.
The results of the comparison agree to within 2 MHz (out of a band of 500 MHz) in frequency and to within 4 dB (out of a dynamic range of 30 dB for the experimental data) in amplitude.
COMPUTATIONAL AND EXPERIMENTAL RESOURCES
The Lawrence Livermore National Laboratory (LLNL) has worked for many years to build unique tools and resources in the areas in EM modeling and microwave/RF antenna design and characterization. It is the unique connectivity at the Laboratory between EM modeling, design, construction, experiments, systems analysis, and field work that allows the lab to work in many leading areas in high power systems. These include the areas of unique EM modeling codes, RF design and component construction, microwave and RF measurements and device characterization, and systems analysis.
Computational Electromagnetic Codes
In support of many individual projects, several unique codes have been developed in electromagnetics. These codes comprise full 2D and 3D time and frequency models which encompass propagation, scattering, dispersive/lossy materials, thin wires, mutual coupling, and complex and arbitrary geometries. These models include NEC(2) [used for decades as the thin wire code in many applications]; AMOS(3) [a 2½D code used for accelerator design, antenna characterization, and radar systems design]; TSAR(4) [a 3D code used for pulse propagation, materials interaction, and antenna design]; and two new codes: DSI3D and PATCH/EIGER which will surpass the capabilities of the existing codes. The important point to note is that the existing codes have many man-decades of development effort, are well tested and verified with theory and experiments(5), and have been used for many broad ranging applications.
The following diagram briefly compares the geometries of some of the codes developed and being used:
Figure 1. AMOS 2½D coaxial cell mesh, TSAR 3D cubical cell mesh, and DSI3D unstructured cell mesh show the differences between the code's geometries.
The type of application drives which of the above codes are used. The NEC wire and surface code is used for large unbounded structures while the AMOS code is used for axisymmetric or 2D problems; especially when complex material models are involved(6). The TSAR 3D code is used for generic problems for pulsed and CW excitation while the DSI3D(7) code is used for highly curved and space filling problems. It should be stressed that unlike other codes, these are general EM codes; the geometry determines which is appropriate for the problem. In all cases, they are Maxwell field solvers (NEC and PATCH/EIGER are frequency domain codes, while the others are time domain codes that can of course be used to solve CW problems).
Experimental Facilities
The Laboratory has experimental capabilities that were used in this evaluation. In addition to the modeling capabilities mentioned above, we also have electromagnetic facilities:Figure 2. The anechoic chamber and it's associated control systems and RF sources can be seen. In this case, a canonical broadband radiating antenna is shown.
COMBLINE ANTENNA VERIFICATION
One of our latest projects includes work done for General Atomics in EM modeling, fabrication, and RF testing of a proposed RF antenna plasma heating system. This modeling was performed using the lab's 3D codes and the experimental characterization was done in the facilities mentioned above. Note that related work as part of other projects has been done in modeling accelerator systems, antenna systems, antenna feed circuits, matching networks, near field antenna effects, and materials modeling which include frequency dependent parameters (e.g., the material properties are functions of whatever frequency components are passing through it). To date, these material properties are isotropic (but frequency varying) but tensor characteristics will be added for upcoming electro-optic and ferromagnetic applications. The General Atomics project will involve most of these technologies to characterize the envisioned system.
For the GA project, the initial antenna EM modeling consists of 4 steps:
In the above list, Step #1 was the most time consuming since it established the baseline antenna modeling performance issues. This step has been completed. To date, Step #2 is being completed now and contains the issues relating to performance and RF energy transport.
The above list was arranged to track the energy flow from the system input to the plasma in a linear modeling progression. It was felt that making modeling and experimental comparisons along the way was important and this exercise has been beneficial. The goal is to model antenna systems that are not practical to fab up and to gain an understanding of the EM effects in the antenna.
For the GA antenna system mentioned in Step #1, a preliminary mockup was made for testing purposes to validate the codes used in the project for this novel configuration proposed by GA(1). The 3D EM model and a photo of the antenna are shown below.
Figure 3. The 3D EM solid model was designed to accurately represent the structure of the antenna.
Figure 4. Photo of the antenna mockup that was actually fabricated.
Earlier versions of the comb line antenna structure were experimentally examined prior to the modeling effort. The mockup used in the modeling exercise was measured extensively in the Laboratory's Electromagnetics facility (mentioned above). The mockup antenna is shown in Figure 5 radiating into the back wall of the facility which is covered by 4 ft. absorbing material. This test setup was used for the measurements made when the antenna cover was not in place and represents the operating configuration of the antenna. The absorbing material was used to remove any room reflections that might occur. As seen in Figure 5, the antenna is fed by coaxial lines and short sections of these lines were included in the 3D EM model. This allowed the well behaved field distribution of the coaxial cable to be modeled and the critical conical feed region was included in the model for both the input and the output ports to the antenna.
Figure 5. The mockup antenna is being tested in the EM lab for modeling comparison purposes. Notice that this experiment did not include the side flanges.
The mockup antenna was constructed at LLNL to make comparisons between the EM models and the experimental results. The results of preliminary models were used to guide the construction of this antenna. The feed region was modified to allow a tapered feed to reduce the input reflection coefficient (see Figure 6) while the strap-to-side wall distance was increased to reduce the amount of residual energy not being coupled to the output port or not being launched. These design changes were incorporated into the mockup design before construction. All results shown are for the mockup that incorporates all of these features.
Figure 6. The feed region of the antenna shows the tapered transition from the coax to the driven strap.
The antenna's input and output performance characteristics are shown in Figure 7 for the antenna in its standard configuration. Models were also generated for experimental comparison purposes for the antenna with/ and with/out the wickets installed and with/ and with/out a cover in place over the antenna. Note that the antenna s operating configuration includes the wickets (shield) and has the antenna radiating into the plasma. The other configurations were for modeling and experimental comparison only. It should be noted that the wickets contributed significantly to the coupled power that actually entered the antenna versus what was reflected back to the input port. These results are shown in Figure 7.
wickets, no lid no wickets, no lid
wickets, lid no wickets, lid
Figure 7. Input (s11) reflection coefficients for the antenna show the increase in power delivery to the antenna due to the increased strap to strap cross coupling via the wickets. The plots are magnitude (dB) versus frequency (100-500 MHz). Note that all four plots have the same scales but for simplicity, not all tick marks are labeled. The s11 blip in the ``no wickets, lid'' case is the cavity resonance of the enclosed box (TE001 mode). All four plots have the same vertical and horizontal scales.
The strap effects can be seen in Figure 8. The actual mechanism is for energy to couple from the input strap to the wicket structures between strap 1 and strap 2. This energy then couples to strap 2 or is launched toward the plasma and the process continues for the remaining energy. The presence of the wickets creates a better input impedance match and thus more energy is delivered to the antenna. The input impedance over the band of operation for the mockup was roughly 50
(from Figure 7). This matched the coaxial drive cable s impedance and thus all of the energy entering the feed cable was delivered into the antenna in this band.
The wave interaction between the input strap and the strap elements is shown in Figure 8. The interaction between the straps and the wickets is also visible.
Figure 8. The wave propagation interaction between the antenna elements is clearly visible as the wave is coupling to Strap 2.
Finally, an estimate was obtained for the energy transport from the antenna and compared to the experimental data (Figure 9). Note that these numbers will be refined as part of Modeling Step #2, mentioned earlier. This data represents a calculation based on the [s] parameters compared against the measured data for the antenna taken in it s radiative mode.
It should be stressed that this represents a comparison of parameters made only at the input ports vs. a measurement made in the radiative region of the antenna. The agreement is very good (5%) considering the assumptions that were made to arrive at the calculated values. Note that the actual comparison of the radiative effect is part of Modeling Step #2.
Figure 9. Energy transport based on experimental and modeling shows the peak in the transmission curve. Note that the peak transport is 24% and 37% (based on the maximum physical area) in the upper curve taken in the experimental facility while the lower curve is based solely on s11 and s21 and has a peak at 18% and 42%.
CONCLUSIONS
In the course of performing the EM modeling, it was noted that the wickets (Faraday shield elements) were acting as coupling mechanisms, that the feed region was important but that the feed drive point was not. Further investigations will be needed to look into tuning the feed for the specific operating frequency to improve performance.
We have illustrated the ability to evaluate, design, fabricate, and characterize a wide range of microwave, RF, impulse, and CW systems that include antennas, materials, circuits, systems, and facilities. We have also described how these previous efforts aided in the modeling and experiments for the RF heating antenna by analyzing the major components of the antenna system in 3D using time and frequency domain codes.
FUTURE WORK
Work will continue on the diagnostics of the antenna with regards to parameters normally seen in tokamak design. We will initially introduce a bulk material and experimentally and through models verify the coupling from the antenna to the material. After having successfully verified the material coupling values, we will introduce a simplistic plasma model into the EM codes.
ACKNOWLEDGEMENTS
The authors would like to thank Cliff Shang, of LLNL, and Dick Freeman of General Atomics, for their past contributions, suggestions, and interest during the course of this project.
Portions of this work were performed under the auspices of the US Department of Energy by the Lawrence Livermore National Laboratory under contract No. W-7405- ENG-48.
APPENDIX
The following are magnified views of one of the blocks from Figure 7 for illustration purposes:
Figure 10. s11 (upper block) and s21 (lower block) show comparisons between EM modeling (solid curve) and experimental results (dashed curve).
REFERENCES
LLNL Disclaimer
UCRL-JC-119968
