Figure 12. Response of the triplexer after applying a final optimization procedure.

On the contrary, convergence of each channel filter, using H-plane symmetry, was obtained with 5 accessible modes, 150 basis functions and 500 Green’s functions modes. The design was performed on a PC with Intel Xeon processor E5-2630 and clock frequency 2.40 GHz. The computational time needed for the analysis of the whole structure was 37 s for 250 frequency points. On the contrary, only 3.5 s were needed for the analysis of one isolated channel filter. This shows that the optimization of the isolated channel filters, which is one of the key steps in the design process, may be very fast.

The design of the structure is performed assuming ideal lossless materials. Once the design is finished, we can make an assessment on the expected losses in the structure. For the calculation of the losses, again the software FEST3D has been used, assuming that the surfaces are silver plated, as commonly done in space applications (with a resistivity of 1.6 10−8 Ω m). The expected minimum insertion losses at the center of the passband are approximately of 1.1 dB for the three channels.

3.2. H-Plane Manifold Multiplexer

In addition to the multiplexer design discussed in the previous section, an analogous strategy can be applied for the design of a similar triplexer but using a manifold composed of H-plane T-junctions. One of the advantages of this topology is that it remains completely 2-D, as shown in Figure 14, including channel filters and the manifold. Consequently, the complete analysis can be reduced to a two dimensional structure, being very e cient, for instance, following techniques such as those described in Perez Soler et al. [2007].

In Figure 14, it can be observed that the orientations of the three channel filters have been judiciously selected so as to fit all filters on the same side of the manifold. The filter of the central channel is the one having more limited space. This is due to the extended width needed in the dual-mode cavities. Other simpler filters such as those implemented in in-line topologies with single-mode cavities will be less limited in space. In any case, if the central filter cannot be fitted on the same side of the manifold, it can always be attached to the other side, at the expense of more footprint.

The response of the designed triplexer is shown in Figure 15. This response has again been obtained after first applying the design strategy presented in this paper and a subsequent final optimization process to correct for small inaccuracies. The level of precision achieved for this second design was very similar than the one obtained for the previous design. Overall, the response is very satisfactory, demonstrating the e ectiveness of the process described in this paper. The dimensions for this second design are collected in Table 8.

PONS ABENZA ET AL. MULTIPLEXER WITH ALL-INDUCTIVE FILTERS 1075

Table 7. Individual Channel Filter 3 Parameter Variation From Initial Manifold Connection to

Final Optimization of the Triplexer

Practical considerations for real applications involve sensitivity issues about the designed component. A sensitivity study has been carried out for the two triplexers designed in this paper. The study has revealed that the E-plane manifold triplexer is less sensitive than the H-plane manifold triplexer. For the E-plane manifold device, a 15 dB return loss can still be expected with errors of 5 μm, while 2 μm are needed to the H-plane case. This shows that tuning screws are needed for a practical implementation.

For the proposed technology, it has been shown that the most sensitive parameters a ecting the response of the channel filters are the resonant frequencies of the resonators. However, sensitivity with coupling elements is much lower. For instance, in Guglielmi et al. [2001] the desired responses of the filters could be recovered adding only screws to tune the resonant frequencies of the resonators in the dual-mode cavities. The tuning could be performed adding one screw in the center of the single-mode cavities, and four screws displaced from the center in the dual-mode cavities, as shown in Guglielmi et al. [2001].

For the tuning of the whole multiplexer, the most critical components are the couplings of the filters attached to the manifold. In this sense, the structure proposed in this work is very convenient, since these couplings are formed with standard inductive windows in traditional single-mode waveguide cavities. Consequently, it is expected that no tuning mechanism is required in these critical coupling elements. If the initial critical couplings are preserved, then the tuning of the internal resonators in each filter can be made with little

Figure 13. 3-D sketch of the designed E-plane manifold triplexer.

Figure 14. 3-D sketch of the designed H-plane manifold triplexer.

Figure 15. Scattering parameters of the H-plane manifold triplexer.

Table 8. Geometrical Parameters in (mm) of the Three Channel Filters After Applying a Final Optimization Process, for the

H-plane Manifold Triplexera

Figure 16. Multipactor thresholds obtained for the most critical filter in E-plane and H-plane manifold triplexers designed in the work (channel 1). Thresholds are shown as a function of frequency inside the passband.

interactions from the rest of the network. Consequently, the tuning of each filter can proceed almost independently, thus simplifying the tuning task of the whole multiplexer.

For space applications, it is also important to assess the behavior of the devices to high power e ects, such as multipactor [Vicente et al., 2005b]. Since the structures proposed are all-inductive, it is expected that multipactor thresholds will be determined essentially by the height of waveguides. Detailed studies have been carried out using the same software tool FEST3D for the two triplexers presented in this paper, using the nominal waveguide height of (b = 9.525 mm) and assuming silver plated walls. For the simulations, an initial population of 500 electrons were used, which usually leads to accurate predictions [Vicente et al., 2005a]. We have verified that in both cases the most critical filter is the one which is closest to the manifold short circuit (channel 1 in Figure 1). In addition, the most critical elements in this filter are the dual-mode cavities. Although multipactor can be triggered in the manifold sections, the thresholds are always higher than in the all-inductive dual-mode cavities. In Figure 16 we present the lowest thresholds obtained for channel 1 filters in both triplexers designed in this paper.

Figure 17. Minimum multipactor thresholds obtained at five selected frequencies, resulting from five experiments conducted with the software FEST3D.