Novel High-Q Partially Air-Filled Pedestal Resonator and Filter Integrated in a Printed Circuit Board (PCB)

In this paper, a resonator and second order filter are fabricated using a novel technological process based on micromachining and thermo-diffusion. The use of this innovative process opens the way to the design of RF components relying on partially air-filled Substrate Integrated Waveguides (SIW). These topologies and particularly the proposed partially air-filled pedestal SIW resonator is suitable to design high-Q, yet compact SIW resonators. In this paper, a study of partially air-filled pedestal SIW resonator is proposed to offer an optimized trade-off between Q-factor and compactness. Then, based on this study, a partially air-filled pedestal resonator working at 5 GHz is designed and manufactured. The measured prototype exhibits a Q-factor of 285, which represents a 53 % increase in Q-factor compared to a fully dielectrically-filled-in pedestal SIW resonator, while the size is kept constant. Finally, a second order filter based on this resonator topology is also designed, measured and discussed.

cult to integrate into systems. 23 The associate editor coordinating the review of this manuscript and approving it for publication was Qi Luo .
In 2001, a technique called Substrate Integrated Waveguide 24 (SIW) was first introduced [1] with the purpose to bridge the 25 gap between planar and waveguide topologies by reaching 26 electrical performances close to those of waveguide struc-27 tures while using planar fabrication processes. SIW consists 28 in designing structures based on waveguide modes, the struc-29 ture being filled-in with dielectric and delimitated by two 30 metal layers that are connected with via hole arrays to form 31 the side walls. As they can be manufactured thanks to clas-32 sical planar technological processes, it results in cheap and 33 light structures that can be easily integrated into systems [2]. 34 Additionally, as the resulting waveguide is filled in with 35 dielectric substrate, it leads to rather compact devices [3]. 36 Nevertheless, although achieved Q-factors with SIW are 37 higher than those reached with standard planar topologies, the 38 use of dielectric materials with non-null loss tangents limits 39 [12], [13], and [14], also called mushroom resonator in [15]   To achieve the realization of these structures, an innovative 87 fabrication process based on micromachining and thermo-88 diffusion assembling was used.

89
The next section details the interest of the proposed topol-90 ogy and the key-aspects of its design while section III 91 describes the manufacturing steps and presents the achieve-92 ment of both a resonator and a second-order partially air-filled 93 pedestal filter. Finally, section IV presents the measurement 94 results and discussion.

96
The pedestal resonator consists of a classical SIW cavity 97 loaded with a metal plate suspended at an intermediate height 98 and short-circuited on one of its sides by one or several metal 99 posts [10], [11]. The cross-sectional view of a pedestal SIW 100 resonator is shown in Fig. 1(a). Height between the bottom 101 ground plane and the pedestal plate is denoted h1 while the 102 height between the pedestal plate and the top ground plane 103 is denoted h2. For practical reasons, the intermediate plate 104 is built upon a dielectric support, which is itself crossed by 105 a via-hole to enable a conductive connection between the 106 pedestal plate and the bottom ground plane. With such a 107 configuration, E-field maximum is located between the top 108 metal layer of the cavity and the metal plate of the pedestal as 109 shown in Fig. 1(b). As mentioned in the introduction, one of 110 the main advantages of pedestal resonators is that they allow 111 for size reduction compared to ESIW or AFSIW cavities. 112 To illustrate the size of the resulting resonator, Fig. 2 shows 113 the top view of a cross-sectional cut of the partially air-filled 114 cavity made in the pedestal plate plane. As denoted in Fig. 2, 115 L is the outside length of the square SIW cavity while W is 116 the width of the pedestal plate. To evidence the interest of the proposed partially-air filled 118 resonator, several eigenmode simulations were performed 119 with different values for the outer area (L × L), Q factor 120 being extracted for each case (Fig.3 (a)). To allow for fair 121 comparisons, the resonant frequency was kept constant and 122 set to 5 GHz. To maintain a constant resonant frequency with 123 various outer area (L × L) values, the pedestal width W had 124 to be tuned accordingly. Fig. 3 (b) shows, for each value of the 125 area (L × L), the corresponding values of the W/L ratio nec-126 essary to maintain a constant resonant frequency of 5 GHz. 127 The simulations were performed with CST Studio Suite with 128 the relative dielectric permittivity of the substrate (ε r ) is 3.55, 129 its loss tangent (tan δ) being 0.0027 and conductivity of the 130 metal (σ ) being set to 5.96 × 10 7 S/m. As a first approach, 131 heights h1 and h2 were both set equal to 0.5 mm. filled configuration to obtain sizes greater than 500 mm 2 .

144
As an example, with the same materials characteristics, 145 the reachable Q-factor of a partially air-filled pedestal SIW 146 resonator (orange lines) is around 300 for a given area of 147 200 mm 2 ( Fig. 3(a)). This Q-factor is computed at a resonat-148 ing frequency of 5 GHz, which is achieved by designing a 149 pedestal with a W/L ratio of 0.55 ( Fig. 3 (b)). For the fully 150 dielectrically-filled counterpart (blue lines), and by consid-151 ering both the same area and the same frequency, which 152 is achieved with a W/L ratio of 0.3 ( Fig. 3 (b)), reachable 153 Q-factor is around 200 (Fig. 3(a)).  Additionally, Fig. 3 shows that partially air-filled pedestal 162 topology also permits to reach higher Q-factors than the fully 163 dielectrically-filled pedestal configuration for the same outer 164 area, thanks to the compensation on the pedestal dimensions. underline that the empty pedestal resonator cannot be manu-168 factured, at least not with classical PCB processes. Indeed, the 169 intermediate conductive plate needs to be supported, although 170 this configuration is unrealizable, it was included in the study 171 for comparison sake.

172
Finally, the partially air-filled pedestal resonator is able 173 to achieve a wide range of sizes and interesting Q-factors 174 (always higher than the dielectrically filled-in structure).

175
From these considerations, and because of its large range 176 of design options and its versatility, it appears clearly that 177 the partially air-filled pedestal topology stands as a com-178 petitive candidate when optimal Q-factor/size trade-offs are 179 seeked. As a counterpart, it necessitates specific features in 180 the manufacturing process, which are detailed in the next 181 section.

III. DESIGN AND MANUFACTURING
Now that the principle and the design of a partially air-212 filled pedestal resonator have been introduced, the structure 213 has to be realized.

214
The conventional way of manufacturing Air-Filled-or 215 Empty-Substrate Integrated Waveguides on Printed Circuit 216 Boards (PCB) processes is to mill a double-sided board that 217 is then mechanically-assembled with copper foils, generally 218 through the means of bolts.

219
Although this results in a cheap process, problems for 220 alignment, reliability, and conductive continuity arise, thus 221 it is not suitable for industrial deployment. For this reason, 222 we choose to fabricate this resonator with a process consti-223 tuted of three stages: micromachining of the cavity, metal-224 lization and thermo-diffusion stacking.

225
Firstly, from a classical copper-cladded double-side dielec-226 tric board (Fig. 5 (a)), a depth-controlled micro-machining 227 or laser-based process selectively removes the substrate and 228 metal to create the pedestal shape, as shown in Fig. 5 (b). 229 In this figure, the left part presents the top view of the 230 resonator while the right-hand side of the figure shows 231 cross-sectional cuts along the AA' and BB' lines. Dielectric 232 substrate is depicted with blue shades, air is white-colored 233 whereas metallization is represented in black.

234
Secondly, via-holes are drilled (Fig. 5 (c)) and the all 235 the surfaces are the metallized, including via-holes, cavity 236 sidewalls and pedestal edges, as presented in Fig. 5 (d). 237 De-metallization of specific areas, such as pedestal or 238 accesses edges is then made by using a precise milling 239 machine or a laser (Fig 5 (e)). Then, a 35-micrometer-high 240 copper foil is bonded, by means of thermo-diffusion, onto the 241 top of the structure to create an enclosed quasi-empty SIW 242 cavity, as shown in Fig. 5 (f). To realize this thermo-diffusion-243 bonded closure of the cavity, a specific metal layer has been 244 deposited on the copper metallization, which creates a strong 245 conductive adhesion with the top copper foil under precise 246 temperature and pressure conditions. Finally, slots are etched 247 on the top copper foil to allow for the realization of planar 248 lines accesses (Fig. 5 (g)).

249
A picture of the realized device is shown in Fig. 6. The 250 next step is to proceed with its electrical measurement, which 251 was achieved with an Anritsu R 3680-20 cell connected to a 252 properly-calibrated Rhode & Schwarz R ZVA 67 Vectorial 253 Network Analyzer.

255
The simulated and measured results can be seen in Fig. 7. 256 A very good agreement, both in the general shapes and in the 257 magnitude levels between simulated and measured results can 258 be noticed. One specific feature of the presented electrical 259 responses is the occurrence of a transmission zero located 260 around 6 GHz related to a direct input-to-output cross-261 coupling. Unloaded Q-factor was extracted from measure-262 ments and the obtained values, Q = 285, which is very close 263 VOLUME 10, 2022 to the simulated one, has to be compared to that of an equiv-   This evidences an unloaded Q-factor increase of 53% for 268 the partially air-filled pedestal topology with respect to the 269 dielectrically-filled counterpart. It is paramount to note that 270 this Q-factor increase was made at no cost regarding the 271 overall volume (14 × 14 × 1 mm 3 ) occupied by the res-272 onator, evidencing the interest of the proposed concept and 273 technology.

274
Further to the manufacturing of a resonator, and to pinpoint 275 the relevance of this topology, a second order filter was 276 designed. An overview of the second-order partially air-filled 277 filter can be seen in Fig. 8.

278
The first step was to characterize the coupling between 279 the two pedestal resonators in order to achieve the desired 280 passband. The inter-resonator coupling is tuned through the 281 variation of D, distance between the central via-holes of 282 the pedestal resonators. The coupling level as a function of 283 the distance D was computed thanks to full-wave simulations 284 performed with CST Studio Suite and the corresponding plot 285 is proposed in Fig. 9. Using this abacus, a second-order filter 286 based on partially-air filled pedestal resonators is designed. 287 The targeted relative bandwidth is 2% and the filter is cen-288 tered at 5 GHz. All the dimensional parameters related to the 289  resonators are similar to those detailed in section III, and the 290 inter-resonator distance D is set to 11.4 mm.

291
A photography of the realized filter is shown in Fig. 10. The    A full-wave simulation with a 2% decrease on the pedestal 310 width (W) and a lowering of the metal conductivity from 311 5.96 × 10 7 S/m to 2 × 10 7 S/m ha lowering of the metal 312 conductivity from 5.96 × 10 7 S/m to 2 × 10 7 S/m s been per-313 formed. Simulation results and measured electrical response 314 are presented in Fig. 12. Characteristics of the realized fil-315 ter together with data from recently-published, size-reduced 316 SIW bandpass filters (including Air-Filled and Empty SIW) 317 in terms of center frequency (fc), order (n), fractional band-318 width (FBW), insertion losses (IL), and surface of the circuit 319 expressed in square wavelengths (computed in the air) are 320 given in Table 1.

321
Additionally, Q-factor extracted from the measurements of 322 the filters responses (from FBW, n and IL using [20]) is also 323 presented. Finally, a Figure