MEMS Accelerometer Design and Simulation
MEMS accelerometers are widely utilized in cars for airbag deployment and in consumer electronics applications such as smart phones. In addition, there is a growing demand for high-end MEMS accelerometers in industrial, aerospace and defense applications such as oil exploration, structural health monitoring for bridges, and inertial measurement units for navigation. To meet these challenges and gain market share, MEMS design teams are adopting new design methodologies that replace costly, time-consuming build-and-test and customized, internally developed design tools. Even teams who have already invested in general-purpose simulation tools are recognizing the need to augment them with additional MEMS-specific tools to accelerate the pace of design exploration and optimization.
Sensitivity, bandwidth, noise, linearity, dynamic range, shock resistance, and temperature stability are all critical specifications that must be satisfied by a MEMS accelerometer design. Going beyond design of a discrete MEMS device, increasing integration presents new design challenges such as cross-axis coupling and parasitic coupling between MEMS and CMOS. Addressing these design challenges while achieving the cost, form factor and time-to-market goals set by today’s highly competitive market requires a new design methodology.
A New Hybrid-Design Methodology
Partnering with leading experts, Coventor has fashioned a new hybrid design paradigm for MEMS accelerometers. A simplified view of the design platform is presented in Figure 2.
The new methodology is demonstrated through an example design, a capacitive accelerometer fabricated with imec’s poly-SiGe MEMS technology, see reference 1. Monolithic integration via this post-CMOS MEMS process yields fewer undesired parasitic effects, better performance, and smaller form factor.
- Rapid Design Exploration
First, a model is constructed in MEMS+ by assembling fundamental elements from its extensive parametric library. The physical 3D model is imported and run in MATLAB, Simulink, or Cadence Spectre for electromechanical analysis and system design. In this design, pull-in voltage was simulated to determine the upper boundry for the drive voltage in order to maximize the sensitivity while maintaining stability. Next, MEMS+ was used to rapidly explore the design space and obtain the preferred resonant frequency or bandwidth for this application.
- In-Depth Detailed Analysis with FEA/BEA
After completing the initial design exploration for a given process flow and deciding on a specific design candidate, designers utilize the field solvers in CoventorWare to do in-depth analyses and finalize other critical details of the design. In this example, the effect of fillets in the beam support on the stress/reliability of the device was investigated. In addition, the impact of the non-ideal anchor structure and etch holes on the natural frequency of the device was also studied. Next, the MemElectro solver in CoventorWare was used to compute the capacitance/electrostatic force of the self-test structure more accurately and contact analyses were performed to evaluate the design of shock stoppers.
- MEMS + IC System Simulation
Using Coventor’s unique methodology, MEMS designers can now include their ASIC and MEMS device together for a full MEMS + IC simulation. The resulting 3D model from the initial simulations is augmented with additional details from the FEA simulations and run in Cadence Spectre. If designers prefer, they can run similar system simulations using the Matlab/Simulink environment, leveraging standard Simulink Toolboxes to optimize the system performance.Co-simulation for this sensor structure and sensing electronics was achieved by importing the MEMS+ model into Cadence Virtuoso. Figure 5 shows the schematic used and the MEMS+ model connected to a simple unity gain voltage amplifier. This provided the designer with an optimal approach to evaluating linearity, sensitivity, and bandwidth of the MEMS accelerometer to be simulated using standard analyses provided by Cadence Spectre. Since the model is fully parametric, MEMS designers can vary all design parameters. Expert users are employing powerful scripting and other automation routines to explore the performance envelope of the accelerometer. Yield analysis is accomplished using Monte Carlo simulation and optimization algorithms are employed to minimize noise and maximize sensitivity.
A Complete Platform for MEMS Accelerometer Design
Coventor’s unique design platform couples best-in-class behavioral, circuit and FEA algorithms for MEMS accelerometer designers to rapidly and accurately explore today’s critical design challenges. The platform offers the ability to simulate the complex multi-domain, multi-physics of the sensing elements and associated electronics, accurately predict noise, develop new packaging strategies, and address manufacturing yield requirements.
The advantages of this new hybrid approach are numerous. Because the underlying physics of the MEMS accelerometers are already included in the MEMS+ models, MEMS designers no longer need to create a reduced order or equivalent model networks from FEA data and/or analytical expressions. This saves valuable time and resources. Because these models are inherently parametric, rapid exploration of design space and process changes can be immediately realized.
In addition, utilizing a platform that employs best-in-class simulators like Cadence Spectre and/or Matlab/Simulink provides MEMS designers the best combination of accuracy, capacity, and speed. As Coventor, Cadence and Mathworks all make speed and capacity improvements to their individual algorithms, users realize an exponential increase in speed and capacity as the improvements from Coventors’ algorithms are multiplied by the algorithm improvements of the others.
L. Wen, B.Guo, L. Haspeslagh, S. Severi, A. Witvrouw, and R. Puers, THIN FILM ENCAPSULATED SIGE ACCELEROMETER FOR MEMS ABOVE IC INTEGRATION.