Optical MEMS Design and Simulation

Optical MEMS or MOEMS have had greatest success to date in digital projection — think DLP technology from Texas Instruments. But MOEMS are increasingly adopted for other applications, including passive-light display technologies, optical networking, focusing and zoom mechanisms for miniature cameras, and scanning. To meet complex design requirements, engineers are adopting a hybrid design methodology that offers significant advantages over home-grown approaches and reduces the need for time consuming and costly build-and-test cycles.

Figure 1: Examples of optical MEMS devices

Design Challenges

The main design challenge for a typical MOEMS device is to optimize switching speed, which involves understanding mode shapes and frequencies of oscillation. Other critical design parameters include contact dynamics, frequency hysteresis, temperature drift, shock dynamics, package deformation and gas damping, and power. To maximize process yield, process and material tolerances also need to be taken into account in the design phase. Furthermore, successful MOEMS design requires the combined expertise of MEMS device and ASIC Engineers, particularly for MEMS arrays where RC delays induce cross talk, reducing screen performance.

Coventor’s hybrid design flow helps MEMS teams to meet their performance criteria while simultaneously reducing cost and time to market. The methodology leverages Coventor’s MEMS+ to enable fast device design and co-simulation with the drive electronics. By enabling the control system to be simulated together with the MEMS device, early in the design flow, costly redesign is avoided late in the design cycle. Layout from MEMS+ can then be passed to Coventor’s Analyzer to access MEMS specific FEM solvers for detailed device design.

Figure 2: Coventor’s novel hybrid methodology for MOEMS design (applied to a scanning mirror in this example)

New Hybrid Methodology

  1. Rapid Design Exploration with High Order Finite Elements
    Users first create the geometric MEMS model in MEMS+ using components from the extensive library of high-order electromechanical finite elements. In the mirror example shown below, mechanical elements model the suspension and gimbal and comb fingers with a circular backbone are used to model the non-linear electrostatic actuation of the mirror and gimbal. The elements use conformal mapping to simulate fringing field capacitance and comb finger out-of-plane fringing force actuation. Because high order parametric elements are used, rather than traditional finite element analysis (FEA), users can assemble and solve hundreds of solutions in the time it takes to run one simulation in FEA.

    Figure 3: MEMS+ high-order components used to assemble a model of a scanning mirror

    The MEMS+ models have been optimized to run in MATLAB/Simulink or Cadence. For an electrostatically driven micro mirror, frequency response, drive angle with drive voltage and transient response are commonly evaluated at this stage. Yield analysis via Monte Carlo simulation provides users with an ability to rapidly evaluate differences in device performance using process and material information provided by their foundry.

  2. Detailed Analysis and Verification with Field Solvers
    The next step is to utilize field solvers to perform detailed analysis of the MEMS device. For the mirror example, Coventor’s Analyzer is employed to extract the damping coefficients for the MEMS+ and investigate the stresses in the suspension beams. To provide additional confidence in the simulation prior to fabrication, MEMS+ and Analyzer solutions can be validated by comparing resonant frequencies.

    Figure 4: Position-dependent gas damping coefficients can be simulated in CoventorWare. These values can be inserted in the MEMS+ model and interpolated for transient simulations of the device.

    Figure 5: CoventorWare simulation of suspension stress distribution and maximum stress with rotation angle

  3. MEMS+ IC System Simulation
    Next, co-simulation of the device with the electronics is done using Cadence or Simulink. This same model can also be used to make an array for a full system analysis of the entire display. An example schematic is shown below. Here, the MEMS+ model array is connected to a driving SRAM actuation circuit and an RC network modeling the line resistance. Because these models are parametric, the designer can quickly and easily adjust the circuit and/or change the MEMS device to explore which strategies yield the best performance.

    Figure 6: Schematic of a mirror array in Cadence and transient simulation results shown in a 2D graph and 3D animation

A Complete Platform For MEMS Design

Coventor’s MEMS+ and Analyzer provide a powerful platform for MOEMS design and addresses the critical design challenges which cannot be solved using general FEM tools. Using this platform, designers may simulate the complex multi-physics of the actuation and sensing mechanisms together with the electronics and MEMS package. Furthermore, Monte-carlo analyses can be used to study and improve manufacturing yield.

This methodology has numerous advantages. First, engineers do not need to hand craft reduced-order or equivalent models from FEA or analytical expressions; all essential physics of the MOEMS are included in the MEMS+ model. This saves valuable time and resources. Second, rapid exploration of process and design variations is possible, because the MEMS+ models are fully parametric and simulate extremely fast compared to conventional FEA models. Third, by utilizing a platform that integrates with best-in-class simulators like MATLAB, Simulink, and Cadence Spectre, engineers have access to the best combination of accuracy and capacity. As Coventor, Mathworks and Cadence all make speed and capacity improvements to their tools, the benefits multiply. Going forward, users can expect simulation performance to exceed their growing demands.

Reference:
Design Optimization of MEMS 2D Scanning Mirrors with High Resonant Frequencies, Ma,W., Chan,H.Y., Wong,C.C., Chan,Y.C., Tsai,C.J., Lee,F.C.S., MEMS 2010

Comments are closed.