Surface acoustic wave (SAW) filters are widely used in cellular and sensing technologies due to their characteristically compact size and low cost. In this article, the design, in-house fabrication and evaluation of SAW filters are described. The effects of interdigital transducers (IDTs) are investigated by fabricating the SAW devices with various parametric variations, including the presence of reflectors, IDT length, separation of ports, the number of reflectors and the number of IDTs per port. The final, optimized, SAW filter demonstrates less than 3 dB insertion loss with greater than 10 dB return loss at 480 MHz without external matching elements. Compared with other methods, simulation with parametric optimization speeds up the design of SAW devices by reducing optimization time.

SAW filters are widely used in sensing1,2 and mobile front-end circuits3,4 due to their compact sizes and high-quality factors.5,6 Since they share the same fabrication processes with the integrated circuit industry, they can be massively produced at low cost as well.7

SAW physics is well-studied and the equations nicely describe the behavior of the filtering response with a simple structure. Traditional modeling approaches include the coupling of modes and the use of Green’s functions;8,9 however, these methods do not depict how detailed design parameters can influence the performance of the filter.

One of the solutions is to use the Finite Element Method (FEM) for handling complex structures. Researchers mainly focus on FEM modal analysis to determine the resonant frequency. However, predicting the frequency response using FEM analysis involves a frequency sweep that requires considerable computational resources. A fully built 3D SAW device requires more than three days for one variation’s computation even with a high-end hardware setting.10

Simplified 2D FEM models are commonly used for resonant frequency prediction. Then the detailed effects of IDT parametric variation are investigated by in-house fabrication and measurement. This article describes the design, 2D FEM simulation, fabrication and measurement of SAW filters. The resonant frequencies of the filters are estimated by FEM simulation. In-house fabrication and measurement are then carried out to investigate the frequency response and the effect of the variation of design parameters. This simulation method speeds up the design of SAW devices by reducing the optimization time when compared to the 3D method.10


SAW Filter Fundamentals

A basic SAW filter has piezoelectric material at its base with IDTs on top. Piezoelectric material induces an electric potential when it undergoes mechanical strain and vice versa. Therefore, the stress-strain relationship and electric displacement-electric potential relationship are coupled together as represented in Equations (1) and (2).11


[D ]is the electric charge density displacement

[e] is the piezoelectric coupling coefficient matrix

[S] is the strain matrix

[ε] is the dielectric permittivity matrix

[E] is the electric field

[T] is the stress tensor

[c] is the stiffness matrix

[et] is the transpose of the piezoelectric coupling coefficient matrix.

Simulation Method

The commercially available simulation software platform, COMSOL Multiphysics, is used to build the FEM model according to Equations (1) and (2). Structural and electrostatic modules are used since the governing equations involve solid mechanics and electric potential.

3D FEM modeling is more reliable than 2D modeling for investigating the influence of detailed features. However, more than one month is required to solve 10 parametric variations in this 3D model even with a simple hardware configuration.10 Instead, the 2D one-port FEM model used by previous researchers is modified and extended to a two-port configuration for analysis.12

Figure 1

Figure 1 Simulated frequency response of the 2D model.

The extended 2D model is chosen over the 3D model to predict the resonant frequency for the SAW filter, which is the most important filter parameter. Simulation time is significantly reduced, and the passband frequency response can be roughly estimated.

The minimum pitch of IDTs fabricated by in-house facilities is 2 μm. As a result, a resonant frequency of around 440 MHz is selected for simulation to satisfy this minimum feature size, and the resonant frequency is also expected to drift slightly. The 2D simulation is easily accomplished within a few minutes (versus several days for a 3D simulation) on a general-purpose laptop computer equipped with an Intel i5 2.4 GHz CPU and 4 GB RAM.

Parameter Variation

Table 1

First, a filter with a frequency around 440 MHz is simulated (see Figure 1). Multiple design variations are then fabricated and investigated. Different parametric variations are listed in Table 1 and Figure 2 shows the 2D structure of the SAW filter with its design parameters. The five selected features are chosen because their impacts on the SAW filter frequency response are relatively vague from the governing equations. Pitch, for example, was not selected as its impact is well-studied and significantly dominates the resonant frequency.

Figure 2

Figure 2 Design parameters.

The first parameter is the presence of the reflector. The reflector comprises a series of connected metal grids with the same pitch as the IDT. Reflectors redirect emitted waves to the receiving port by reflecting or absorbing the incoming waves, so a filter equipped with reflectors is expected to perform better. The second parameter is the length of IDT overlap. Overlaps of 20, 100, and 200 μm of are selected for fabrication. An inefficient overlap deteriorates bandpass performance and excessive length wastes wafer area. The third and fourth parameters are the number of IDTs per port (N) and the number of reflectors per port (M), respectively; 10, 20, 40 and 80 pairs of N or M are selected. The last parameter is the separation of the two ports; 100, 200, 400, and 800 μm of separation are used. A shorter separation is expected to have less loss.

All variations are expected to affect the resonant frequency and bandwidth slightly, but the effects from variation of these five parameters are expected to be significantly less than the pitch. When the pitch size is fixed by the fabrication technique, however, tuning the other design parameters becomes important. The values shown in bold text in Table 1 are used when the other features are varied. For example, when the length of IDT overlap is varied from 20 to 200 μm, all the SAW filters have reflectors, 40 IDTs per port, 40 reflectors per port and 200 μm separation between the two ports.