MEMS Structure Extraction via Segmented Beam and Spring Modeling

The patent segments the complex MEMS design extraction process into distinct stages: identifying basic MEMS elements (beams and supports) from layout data, deriving intermediate MEMS objects from these elements, and finally deriving output MEMS objects (mass and spring models) from intermediate objects. This multi-level segmentation allows standard EDA tools to handle each simpler sub-task while collectively achieving accurate MEMS structure extraction that would be impossible with a single monolithic approach.

The patent introduces intermediate MEMS objects as a mediating layer between the basic MEMS elements identified from layout data and the final output MEMS objects used for modeling. These intermediate objects serve as a bridge that transforms geometric layout information into physically meaningful structural models, enabling accurate representation of MEMS devices while working within the constraints of existing EDA tool capabilities.

The patent performs preliminary extraction and modeling of MEMS structures during the design phase, creating accurate mass and spring models before fabrication. By deriving output MEMS objects and generating models early in the design process, the system enables virtual verification of device performance, allowing designers to identify and correct issues before committing to expensive and time-consuming physical fabrication and testing cycles.

The patent creates digital copies of MEMS structures through extracted geometric data and generated models that replicate the physical device's behavior. These virtual models serve as substitutes for physical prototypes, enabling performance verification and optimization in silico, thereby reducing the need for multiple iterative fabrication and testing cycles with actual hardware.

The patent divides the complex extraction task into hierarchical levels: first identifying simple geometric beams and supports from layout data, then deriving intermediate objects that capture structural relationships, and finally generating output objects that represent physical properties. This segmentation transforms an intractable single-step extraction problem into a series of manageable processing stages, each with lower computational complexity while maintaining cumulative precision.

The patent applies different levels of extraction detail and processing complexity to different regions and types of MEMS structures based on their specific characteristics. By deriving intermediate objects with locally appropriate levels of detail and applying context-sensitive derivation rules, the system achieves high precision where needed while avoiding unnecessary computational overhead in simpler regions, thereby balancing accuracy with processing efficiency.