By Gunar Lorenz, Director, System Level Simulation
Once again we are entering the final phase of a MEMS+ release cycle. We are tying up loose endsfor for another exciting release, over the summer, of our unique MEMS design software. I believe that the results of our latest research and development efforts will impress our users. You may ask: what’s so exiting about the new release? Well, it will surely require more than one blog to tell you about the many new features. For this blog, I will focus on the new capabilities in MEMS+ 5.0 that address the special design challenges presented by micro mirrors. Even if you are not working on micro mirror applications yourself, you may have heard about the new mini-projectors for smartphones from Fraunhofer or the new industrial devices from companies like Mirrorcle Technologies, Inc., Hamamatsu or Preciseley Microtechnology Corp:
If you haven’t heard of these, be assured that the actual list of companies working on new mirror designs is much longer.
Engineers tasked with creating these new micro mirror devices are confronted with meeting stringent design specs for switching speed, frequency stability, temperature stability, shock resistance, power consumption, etc. To ensure that their design meets these specs, they must be able to simulate the full multi-physics device itself and its surrounding electronics. This is only feasible with a very efficient model of the micro mirror. Classic modeling approaches based on hand- crafted reduced order models fail for mirror designs due to the complex physics involved. To be candid, some of the physics left us scratching our heads for quite a while.
First, typical scanning and projection mirrors are driven electrostatically far outside of their linear regime, leading to frequency hysteresis. Electrostatic comb structures, which tilt the mirror out-of-plane, disengage completely during the mirror operation. At the end, there are only fringing-fields left to pull the mirror back to its horizontal position. Calculating the correct electrostatic finger forces is already a big challenge for conventional field solvers. We on the MEMS+ development team needed to beef up our existing comb library components to make sure they would still be accurate even with the comb fully disengaged. It was apparent that the classic plate capacitor approximation would not do the trick! But ultimately we were successful, as shown in the graph below, by the excellent agreement between MEMS+ with disengagement and our CoventorWare field solver.
Second, our MEMS+ comb models didn’t support comb fingers between two moving flexible structures, a clear modeling requirement for gimbaled two-axis mirrors. That, too, is now possible in MEMS+ 5.0, as shown in the models below:
MEMS+ 5.0 comes with a new set of comb finger models that allow users to create very accurate, parametric multi-physics mirror models for fast transient simulations in MATLAB, Simulink, and Cadence Virtuoso. How fast? Well, Coventor’s development team knows all too well that even fast simulations are never fast enough. Our users continually pose new simulation challenges, and micro-mirror designers are no exception! They asked us to improve the transient simulation speed to allow them to run frequency hysteresis curves in MATLAB, as shown below:
Simulating frequency hysteresis requires transient simulations with many, many time steps. The large number of time steps required pushes simulation times, even for MEMS+ models, into the range of many hours. Clearly longer than a coffee break! To cope with the challenge, we needed to find new ways to speed up our models. Parallel processing and further optimizing our mathematical algorithms would no longer do the trick.
We found our way out in the form of reduced order modeling (ROM). Our new MEMS+ ROM export capabilities are another of the big highlights of our coming 5.0 release, but that is a story for another blog. Stay tuned!