“Design over temperature” — New approach to improve LSAW filters thermo stability

Poor temperature properties of LSAW (Leaky Surface Acoustic Wave) devices are one of the principal problems of applying these devices in mobile communication systems. The TCF (temperature coefficient of frequency) of BAW (Bulk Acoustic Wave) filters, for example, are at least twice better compared to LSAW devices, so LSAW filters on 36-48 LiTaO₃ should have much steeper transition bands for the same specification requirements. This causes increase of the LSAW filter IL (insertion loss) and cost due to bigger die size and inferior yield. The problem of improving temperature stability of SAW devices has been discussed in many papers for years. Many possible solutions were described in [1]. Currently, there are two main approaches to improve LSAW filters TCF [2], [3]. First one is based on deposition of a thin dielectric layer having positive TCF onto the filter working surface. The second approach uses bonding of a thin piezoelectric wafer onto a thermo stable carrier. Both methods can improve filter TCF by changing the effective thermo expansion coefficient and LSAW velocity. Unfortunately, both methods create new problems such as increased cost and deteriorating performance and/or reliability. In this work we propose a novel approach to improve TCF. It is based on changing the device impedance over the working temperature range and does not require a manufacturing technology change. We call this approach "Design over Temperature" (DOT). The idea is to design a filter with a special shape of impedance, improving the passband edges for varying temperatures (improving low edge under "cold" conditions and high edge- under "hot"). This new approach is based on two components: -model parameters (COM (coupling- of- modes) model in our case) are characterized not only for ambient temperature 'ta' but for edges of the operational temperature range (tmin=-30°C and tmax=+85°C) too; -the design optimization process is organized for three sets of model parameters in parallel relative to full temperature range of the filter specification requirements. The goal of the optimization process is to improve the worst from three performances out of "cold", "normal conditions" and "hot" and not just the room temperature performance for every specification requirement. First, we have extracted COM parameters for our fabrication process for three different temperatures: t_a, t_min, t_max and described temperature dependencies for all of them. The large difference in TCF for left and right frequency response slopes is trivially demonstrated using this data. Second, we setup a single optimization process for three different sets of model parameters. We have designed an RX filter for PCS (and UMTS Band 2) application based on the proposed approach. This filter has better IL over working temperature range, steeper left transition band and twice better effective TCF than filters designed by conventional design process.