A contribution towards model-based design of application-specific MEMS

Highlights

• A model-based design and synthesis flow for MEMS accelerometers.

• Automated generation of sensor designs and models for co-simulation and layout data.

• Development of an adaptive readout ASIC for validation measurements.

• Validation of the method by measurements of a synthesized accelerometer.

Abstract

The design process of heterogeneous systems containing electro-mechanical components and electronic circuits involves expert knowledge, methods, and tools from different engineering domains. Cost-efficient research and development of such heterogeneous systems requires a systematic design flow without gaps. A contribution towards this global goal is presented in this article. A development and synthesis tool for one-dimensional accelerometer MEMS has been implemented, calculating sensor solutions and generating the models and layouts required for a hierarchical design flow in an automatic, module-based approach. Utilizing this flow, different accelerometers have been designed, manufactured, and characterized. A dedicated readout ASIC was developed to validate their dynamic behaviour.

Introduction

Microelectromechanical Systems (MEMS) are important drivers of future product development in many market segments with quantities and revenue in billions. They consist of mechanical components like sensors and actuators in combination with readout and control electronics on just a few square millimetres of silicon. Due to their small sizes and low costs, MEMS are key components for the Internet of Things, where they form a basis of the interface between the virtual and the real world [1]. MEMS development entails a growing demand for support by design methods and models. These should enable handling the ever-growing complexity of system design and the incorporation of its multi-physical character. The overall aim is to tightly interleave, systemize, and partially automate the design methodology of electro-mechanical and electronic components in order to enable access to application-specific MEMS development, not least for small and medium enterprises.

The goal of this work is to introduce the showcase implementation of a LEGO principle of the kind that is available in application-specific integrated circuit (ASIC) development to enable the fabless design of multi-physical systems. Such a principle yields lower development time and complexity as well as reduction in costs, compared with full-custom system design. At the same time it enables higher flexibility and more variety of products, meaning that designing for small numbers of pieces becomes more profitable, which opens market opportunities and possibilities of new applications. Thus our design method gets its bearings from modern ASIC development that combines analog with digital circuit design and decouples circuit design from the development of semiconductor technology.

In ASIC design, a model-based top-down strategy with different hierarchy levels is used, starting with a specification and generating models that are iteratively refined towards the layout level. The right-hand side of Fig. 1 illustrates this approach. Simulations on both system and device level are performed to adapt the functionality of a design to the specification. A kernel principle of this process is the usage of primitive devices provided by the foundry along with accurate, mostly lumped models in a process design kit (PDK).

For a model-based design strategy of the overall system co-simulations of electromechanics and electronics are required. Hence the ability to embed mechanics in a typical IC design environment is beneficial. The benefit, not only from the automation aspect, is that mechanics and electronics can be simultaneously considered, which may include shifting functionality from the electronics to mechanics and vice versa. As an example, a mechanical performance may be relaxed which might be compensated by a higher performance or different specification of the electronic part.

Commonly used tools for the design of the electro-mechanical parts of a sensor are FEM simulators like ANSYS. New system-based design methods ([2], [3], [4], [5]) were developed in the late 90s, which led to the publication of commercial tools like MEMS+ (Coventor), MEMS Pro (SoftMEMS) and SYNPLE (IntelliSense). These tools and methodologies can be utilized to design the electro-mechanical layout of MEMS, but they can as well be used in an integrated design flow of electromechanics and electronics, enabling interactions at different hierarchical levels and thus allowing system simulations of the whole MEMS sensor.

Although the idea of embedding mechanics in typical IC design environment is fairly old, this topic is still subject to intensive research and development activity. Recent contributions within the scientific community include the investigation of packaging aspects using mixed-domain [6] and mixed-abstraction-level simulation [7], yield optimization using statistical methods [8], and even the development of new co-design/simulation platforms for heterogeneous systems [9]. All these approaches close the gaps between the different simulation tools, and thus solve the problems caused by missing interfaces. However, none of them supports an automated synthesis. The basic architecture of the MEMS has to be created by hand or selected from a library. All the tools and methods optimize the parameters of a given topology, but the task of selecting an appropriate topology of the MEMS is left to a human.

The presented development and synthesis tool for one-dimensional accelerometer MEMS combines several commercial tools and assists designers with an automatic, module-based determination of the basic structure, i.e. topology of the electromechanics, and the geometrical parameters of the electro-mechanical sensor part. Basically, the tool generates all the models required for a structured design flow of the electromechanics as well as the layout automatically (Fig. 1, left-hand side). The generated models allow an interaction between the design flow for the electronics and the one for the electromechanics (Fig. 1, connection of left- and right-hand side).

All steps on the left-hand side of Fig. 1 will be illustrated starting from basic structure generation followed by sizing, variant generation, and simulation down to the sensor layout generation, as described in Section 2. Section 3 covers the design of the readout circuit that was needed to characterize the developed accelerometers. In Section 4, we verify our approach with measurement results of a synthesized accelerometer using the readout ASIC. A final discussion in Section 5 concludes this article.