A Fully Parameterized Fem Model for Electromagnetic Optimization of an RF Mems Wafer Level Package

Phone: +39 (0)51 20 93049
Fax: +39 (0)51 20 93779 ABSTRACT In this paper we present our efforts in characterizing and optimizing the influence of a Wafer-Level Packaging (WLP) solution on the electromagnetic behaviour of RF-MEMS devices. To this purpose, a fully parameterized FEM model of a packaged Coplanar Waveguide (CPW) is presented in order to optimize all the technology degrees of freedom (DoF’s) made available by the fabrication process of the capping part. The model is implemented within the Ansoft HFSSTM electromagnetic simulator, after its validation against experimental data. Moreover, a simulation approach of a capped RF-MEMS varactor is shown. It is implemented in the Spectre© simulator within Cadence© environment. The MEMS part is treated by means of a compact model library implemented in VerilogA© language. A lumped elements network accounting for the parasitics surrounding the intrinsic RF-MEMS varactor is extracted from experimental data. Finally, the S-parameters description of the package, obtained by Ansoft HFSSTM simulations, is included in the Spectre© schematic.

1. INTRODUCTION Packaging has recently been identified as the enabling factor of electronic system performance enhancement and consequently, its technology has gained considerable attention [1]. Concerning MEMS devices, the packaging plays even a more critical role. Indeed, since MEMS devices contain movable parts, like very thin suspended membranes, they need appropriate protection. Factors like shock, moisture and dust particles can partially or totally compromise the proper functionality of such devices.
   Moreover, when dealing with MEMS for Radio Frequency (RF) applications, additional issues related to the packaging come in [2]. For encapsulation of MEMS structures a protective substrate is usually employed (Wafer-Level Packaging). This additional part introduces parasitic (capacitive and inductive) effects related, for example, to vertical through-wafer vias. In addition, after the wafer-to-wafer bonding, the capping part is very close to the device substrate. Hence, the reduced air gap causes electromagnetic couplings between devices that introduce additional losses and mismatch. Parasitics introduced by the application of the capping part have to be reduced as much as possible in order not to compromise the RF functionality of the packaged MEMS devices.
2. ANSOFT HFSSTM VALIDATION Before exploiting Ansoft HFSSTM to optimize the package design with respect to its electromagnetic behavior, this has first to be validated against experimental data. To this purpose, simulated results of capped 50 Ÿ CPW’s and shorts have been compared with measurements. The package fabrication is based on the etching of through-wafer vias subsequently filled with Copper and is provided by the DIMES Technology Centre [3].
   Frequency [GHz] 0 2 4 6 8 10 12 14 S11 [dB] -60 -50 -40 -30 -20 -10 CPW (Experimental) CPW (HFSS) Fig. 1: Comparison of the experimental and simulated S11 parameter for a capped CPW. ©EDA Publishing/DTIP 2007 ISBN: 978-2-35500-000-3

J. Iannacci, J. Tian, R. Gaddi, A. Gnudi, and M. Bartek A FULLY-PARAMETERIZED FEM MODEL FOR ELECTROMAGNETIC OPTIMIZATION Frequency [GHz] 0 2 4 6 8 10 12 14 S21 [dB] -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 CPW (Experimental) CPW (HFSS) Fig. 2: Comparison of the experimental and simulated S21 parameter for a capped CPW. In Figures 1 and 2 the measured and simulated S11 and S21 scattering parameters for a capped CPW are compared, showing good agreement. For instance, the offset between the two curves for the reflection parameter (S11) is 5.3 dB at 12 GHz (Figure 1). On the other hand, the offset between the simulated and measured S21 parameter is 0.14 dB at 6 GHz. The CPW is 1350 ȝm long. Signal and ground lines width is 116 ȝm and 300 ȝm respectively and the gap is 65 ȝm. Package thickness is around 280 ȝm and through-vertical vias diameter is 50 ȝm. S-parameters for other packaged CPW’s and shorts topologies are similar to the ones shown in previous plots. This confirms Ansoft HFSSTM to be a suitable tool for the accurate prediction of capped structures RF behaviour. 3. CAPPED LINE PARAMETERIZED MODEL When dealing with several technology degrees of freedom (DoF’s), their parameterization allows for fully automated optimization. In this work, we focus on the parasitics reduction of the package applied to 50 Ÿ CPW’s and shorts instead of actual RF-MEMS devices. This choice is done mainly because the influ

…(Full text truncated)…