MEMS design

The microelectronics industry is entering the era of Cloud Engineering Simulation to slash the costs and risks of new technology development and speed time-to-market in spaces like semiconductors, MEMS sensors, RF front ends, biomedical and driverless cars. In the run-up to SEMICON Europa, 12-15 November, 2019, in Munich, Germany, SEMI spoke with Ian Campbell, CEO of OnScale, about the new paradigm of Cloud Engineering Simulation. Campbell shared his views ahead of the SMART Design Forum, 14 November, 2019, 14:30 to 17:00, in Hall B1, TechARENA 1 at SEMICON Europa. Registration is open. Join the forum to meet experts from OnScale and other key industry influencers. Attendance is free of charge for all SEMICON Europa visitors.SEMI: How did your adventure with OnScale start?Campbell: I’m an engineer. When I was still in high school, I took a night class at Nashville Tech to learn AutoCAD R14, and I’ve been designing and engineering things ever since. I was introduced to Desktop Simulation in my bachelors of mechanical engineering program and used many types of simulation tools for massive design studies at the Aerospace Systems Design Lab at Georgia Tech. I’m a simulation junkie.I started my first Silicon Valley high-tech company, NextInput, in 2012 with Dr. Ryan Diestelhorst (now VP of Strategy at OnScale), to commercialize new ForceTouch and 3D Touch technologies based on our patented MEMS force sensors. At NextInput, we bought hundreds of thousands of dollars of engineering software, but were always frustrated by slow, inaccurate engineering simulation results. We dreamed about running massive simulations on Cloud Supercomputers and creating true Digital Prototypes that could replace costly, time-consuming, and risky physical prototypes.When I got the chance to join the team that became OnScale in 2017, I jumped at the opportunity. At OnScale, we took engineering simulation solvers that had been developed for the U.S. military to run on U.S. Department of Defense and DARPA supercomputers and built a cloud supercomputer platform on Amazon Web Services to run the solvers. The net-net is the world’s first on-demand, infinitely scalable Cloud Engineering Simulation platform. Now, we routinely run massive multi-billion degree of freedom simulations for Fortune 100 companies, including many from the semiconductor and MEMS industries. Since our business model is to charge per core-hour for simulations, the incredible capability we built is cost-effective and available to small startups as well. SEMI: How is the semiconductor design ecosystem evolving? How is Cloud Engineering Simulation applied to semiconductor and design industries?Campbell: The entire industry is experiencing a massive acceleration in product launch cycles and increased competition. New markets like IoT and 5G are reducing semi/MEMS product cycles from years to months. That, in turn, puts enormous pressure on semiconductor and MEMS designers. Missing a key product introduction like a flagship smartphone launch can literally make or break a company.A reliance on traditional engineering methods – schematic capture and layout of a chip, taping out (physically prototyping the chip), performing engineering validation on an e-bench, qualifying the chip (or not qualifying it and going back to the drawing board), and finally launching mass production – is no longer sustainable from a competitive perspective.Instead, market-leading firms are turning to Cloud Engineering Simulation and Digital Prototypes to explore massive design spaces, find optimum designs that beat the competition in every KPI (size, power, performance), and digitally qualify designs before ever cutting silicon, ensuring that designs are robust over their intended operating environments and performance envelopes. Large thermal analysis of a chip on a circuit board executed quickly on the OnScale Cloud Simulation Platform SEMI: Can you give us an example? Campbell: A great example is thermal analysis. Thermal effects have always had huge impacts on MEMS device performance and, more recently, they are beginning to impact performance of next-gen semiconductors, especially GaN power electronics for electric vehicles (EVs).Conducting a full system-level thermal analysis of something like an EV power management system – a power IC in a package, on a board, in an enclosure, under various loading conditions – has been a challenge from a simulation complexity perspective (degrees of freedom) and from a parametric sweep perspective (running hundreds or thousands of simulations to optimize chip placement, routing, etc.). To run these sets of simulations using legacy desktop simulation would take weeks, perhaps even a month or more. To run these massive simulations in parallel on cloud supercomputers using OnScale takes days or even hours.Our customers routinely run very large simulation studies on OnScale Cloud for thermal simulations, RF filter simulations, MEMS simulations, packaging simulations (what we call Digital Qualification), and many more use cases.SEMI: What’s one of your strategic objectives for 2020? Campbell: For 2020, we’re doubling down on MEMS and semi simulation capabilities. We will be launching additional solver capabilities like EM that will be critical in our strategic markets like 5G. We will also be launching a Cloud API so that engineers can integrate OnScale directly into their existing engineering workflows (e.g. MATLAB or EDA/CAD tools) with just a few Python commands.SEMI: Can you share one prediction for the future of semiconductor design solutions? share?Campbell: I think we will continue to see MEMS and semi designers push the envelope and bring smaller, more performant, more cost-effective solutions to market. I’d like to see more highly cost-effective flexible semi/MEMS designs come to market to enable next-gen IoT and IIoT applications. I’d also like to see more biomedical applications – biomems, microfluidics, and labs on a chip for all sorts of life-enhancing applications.SEMI: What are your expectations regarding the SMART Design Forum at SEMICON Europa 2019 in Munich? Campbell: I’m looking forward to getting back to my roots in MEMS/semi design and chatting with other designers about the future of engineering and the future of semi! Ian Campbell is a twice venture-backed Silicon Valley CEO and expert in MEMS sensors, semiconductor technology, and engineering software. Most recently, Ian co-founded OnScale, a Cloud Engineering Simulation startup backed by Intel Capital and Google’s Gradient Ventures. OnScale is revolutionizing engineering by combining world-class multiphysics solvers with Cloud supercomputers, machine learning, and artificial intelligence. Prior to co-founding OnScale, Campbell served as founder and CEO of NextInput, where he led the startup through multiple rounds of funding – totaling $12 million and an additional $4 million in research contracts with government and industry partners – and built a world-class team of engineers and scientists who developed 3D Touch and ForceTouch technologies for smartphones, wearables, industrial, and automotive interface applications. He also secured the first major smartphone OEM design wins in Asia. Campbell earned his B.S. in mechanical engineering from Middle Tennessee State University, and his MSAE in aerospace engineering and MBA from Georgia Institute of Technology.Serena Brischetto is senior manager, marketing and communications, at SEMI Europe.

Creating a custom Internet of Things (IoT) IC is challenging because it involves multiple design domains (digital, analog and RF). Creating a sensor-based IC that combines electronics that use the traditional CMOS IC design flow with a MEMS sensor on the same silicon die, however, can seem impossible. Couple the co-design and verification challenges with a lack of traditional process design kit (PDK) support for MEMS, and you have a tough road to travel to get your IoT designs to market.What can we do to make the sensor-based IoT design community successful?Understanding the ChallengesThe sensor-based IoT IC typically features a MEMS sensor (and optional actuator) that interact with the real world. Analog and digital circuitry processes the signals and sends them to a CPU. The CPU provides the “smarts” to process the data from the sensor and then sends processed data via a radio to the Internet; alternatively, the CPU could activate the actuator. A typical sensor-based IoT IC (Source: Mentor: A Siemens Business) Based on the complexity of the system, designers face many co-design challenges: Analog design requirements imposed by MEMS: MEMS devices often require high voltages and multiple power supplies; they emit small signals that need amplification and conditioning; and they are sensitive to the environment and require calibration. Design flow interactions: Parasitics from MEMS devices might affect circuits and vice versa. Circuit designers need MEMS models for impedance and timing. Integration: MEMS devices operate at different timescales than circuits, which adds a layer of complexity. Compounding the problem is a lack of MEMS PDKs and methods to tie together ICs and MEMS PDKs for integration and cross-verification. After conquering the co-design challenges, the design team has to address mixed-domain simulation challenges that include: Simulating the system: This requires verification of MEMS, digital, analog and RF circuitry with embedded software that runs on the CPU. Timescales: These vary widely, from a single deflection of the MEMS transducer in femtoseconds to a seconds-long simulation of the embedded software performing a measurement and transmitting data. Simulation time: Simulation of a behavioral digital design is extremely fast. However, the system simulation requires stand-in models that incorporate the behavior of the analog and MEMS block to simulate in an acceptable amount of time. The challenge of timescales for co-simulation. (Source: Mentor: A Siemens Business) MEMS is the KeyThe reality is that it’s the MEMS device that adds extra complexity to the sensor-based IC design and verification flow. To amplify the problem, the MEMS manufacturing process is not nearly as mature as the standardized IC process. For example, the standardized IC process includes ready-made PDKs that include everything designers need to move through design and verification flows. Foundries often provide soft and hard IP to quickly build-out design, and EDA tools provide high levels of automation enabled by abstraction and a standardized IC flow. How will MEMS-based design evolve?MEMS-based design must catch up to the standardized IC process. The first step is providing MEMS PDKs that include: Multi-physics domain design rules and material properties Packaging information Wafer and bonding information Fabrication information We must also tackle issues associated with these PDKs, including: Ownership, distribution and maintenance of the PDKs Consensus on the contents of the PDKs Merging of CMOS and MEMS PDKs The industry needs to move toward standardized MEMS manufacturing processes with available PDKs. Companies must provide IP and recommend structured design methods for co-design and verification of ICs that incorporate MEMS. How can EDA help with these flows?The EDA ContributionEDA companies must work with teams in the MEMS IC co-design space, collaborating with MEMS fabricators to help enable PDKs. By incorporating PDK support within their own tools, EDA companies can provide an integrated custom IC flow that allows teams to design and verify MEMS-based ICs. For details about this flow, click here to download the Mentor whitepaper: Fusing CMOS IC and MEMS Design for IoT Edge Devices.Greg Lebsack brings 25 years of executive and technical management experience — along with a proven track record of building strong teams and delivering predictable results — to his role as general manager of the ICDS division of Mentor, a Siemens Business. Lebsack joined Mentor in 2015 after that company acquired Tanner EDA, where he was president. Prior to Tanner EDA, he held management and technical positions in a number of different industries and companies, including Sprint, General Electric and McKinsey Co. Greg holds a bachelor’s degree in business administration from Northern Arizona University.Greg Lebsack recently presented on the topic of Integrated Co-design of MEMS/IC at the MEMS Sensors Technical congress, a technical conference organized by the MEMS Sensors Industry Group.