Part 1 described Janusz Bryzek‘s ambitious goal of a $1 trillion market for microelectromechanical systems (MEMS) that was the focus of the MicroElectronics Packaging and Test Council (MEPTEC) 10th annual MEMS Technology Symposium. In addition, sensor swarms, road mapping and market numbers were covered. Challenges, example applications, and key takeaways are discussed here along with a final score card on the $1 T market.
The three challenges – simulation, device packaging, and test methods – Janusz raised as limiters to growth of MEMS sensors were addressed in two presentation:
Stephen Breit (Vice President Engineering of Coventor) in his presentation “Realizing the Full Potential of MEMS Design Automation Software” discussed the design challenges of MEMS devices. MEMS companies have recently started to move from very proprietary manufacturing processes to larger scale manufacturing to satisfy consumer product demands of extreme high volume at low costs. To further decrease costs there will be a transition to extremely high volumes concentrated at a few manufacturing facilities including the rise of fabless MEMS, as described by JC Eloy in his presentation (detailed in Part 1).
In order to reduce the design cycle and external costs, it is essential to be able to simulate both the final design and the fabrication process. Through such simulation the number of design iterations will be reduced along with the associated cost of each prototype run. The simulation tools need to understand both mechanical and electrical designs and be able to simulate process variations. The ultimate goal is to simulate all the specifications on the product data sheet. Coventor has started to integrate their different design tools to enable both the mechanical design and analog/mixed signal electrical design of the MEMS device to be simulated at the same time. In addition, like the Physical Design Kits (PDKs) for complementary metal-oxide-semiconductor (CMOS) manufacturing the MEMS foundries will need to develop MEMS Design Kits (MDKs) which describe their processes and material properties in great detail. These MDKs will then be used by the simulation tools to calculate the proper performance.
In “High Volume Assembly & Test Solutions to Meet the Rapidly Growing MEMS Market”, Russell Shumway (Sr. Manager, MEMS & Sensor Packaging, of Amkor Technology) discussed the wide array of materials and processes they support in test and packaging (the “back end”) for both semiconductors and MEMS sensors. In the last twenty plus years there has been a proliferation of package types driven by the diverse set of applications within the MEMS market. All of these MEMS devices have had different test requirements too: pressure sensors need to be pressurized, accelerometers need to be accelerated, gyroscopes need to be spun, temperature sensors need to be heated and cooled, etc. Therefore the test processes need to accommodate all the different device requirements and package configurations to make sure devices are fully tested and are within specification when shipped. These challenges haven’t been specific to Amkor but to all outsourced assembly and test (OSAT) suppliers. However, unlike others, Amkor has invested heavily to solve these particular challenges including developing strip test to lower the cost of testing MEMS devices. Mr. Shumway also indicated that as volumes of MEMS devices continue to rise over the next ten to twenty years, the will be a need for greater standardization in processes and test in order to meet customers’ cost targets.
In “CeNSE: Awareness through A Trillion MEMS Sensors”, Rich Friedrich (Director of the CeNSE Program for Hewlett-Packard) went into detail about HP’s sensing solutions. HP is emphasizing the total solution – which is not just the sensor but how to make a measurement, interconnect the sensors to data processing equipment, manage large volumes of data, and perform analysis on the data to produce actionable results. I.e. people and businesses value the knowledge not the raw data. A prime example of this is how they build large scale networks of up to one million sensors and perform the analysis for oil exploration by Shell and others. He also provided several interesting examples of how very large datasets are visualized.
The final session of the conference, “MEMS for All of Us”, consisted of product demonstrations that was introduced by Tristan Joo (Board Director & Co-Chair of Mobile SIG, Wireless Communications Alliance). Mr. Joo started with a presentation “Fusing Sensors into Mobile Operating Systems & Innovative Use Cases” describing how sensor fusion is enabling mobile applications. In addition, he reviewed the current state of MEMS sensor applications in the mobile market. Sensor fusion is the synergy that results from combining the data from multiple sensors of different types along with algorithms and contextual data to produce higher quality results (more accurate, more repeatable, more predictive of intent, etc.). These results are then used as triggers or inputs to application software.
Today’s smartphones typically have twelve to eighteen sensors each with at least five of these sensors being MEMS based. However, less than one half of a percent of all applications in the iPhone and Android application stores uses these MEMS sensors in a meaningful way. This is due to a lack of developer knowledge about how to best utilize and develop software for these sensors. Fragmented hardware, confusing software infrastructure, and too many different operating systems all make it difficult to use the existing sensors. But as the mobile operating systems mature, more support will be built in to enable the developers to implement new and innovative solutions driven by the wide array of sensors.
After Mr. Joo’s introduction, each company demonstrating a new product using MEMS sensor technology provided a brief introduction to their company and products:
- HillCrest Labs – Justin Zink (Sales, Marketing and Business Development) demonstrated their natural motion based user interface in the form of a television remote control. The user moves the remote in a variety of gestures to control the system instead of pressing buttons. For example to indicate “okay”, you move the remote like you were making a check mark in the air. They sell sensor modules, Freespace MotionEngine software (which interprets the sensor data to determine the meaning of the gesture), and complete remote controls. They provide he remote controls both as an original equipment manufacturer (OEM) and as direct to consumer sales as their own Scoop product.
- Movea – As a spinout of CEA-Leti, Movea develops intellectual property (IP) cores and supporting infrastructure for mobile applications. Dave Rothenberg (World Wide Marketing and Communications) described how they are using their years of MEMS sensor design and fabrication expertise to develop their SmartMotion suite of products. This suite includes motion IP cores and embedded software purchased by other companies to include in their own application specific integrated circuit (ASIC) designs. Software development tools and end applications are also include in this suite.
- Sensor Platforms – Kevin Shaw (Chief Technology Officer) described their FreeMotion Library which integrates sensor data to allow mobile devices to become “Context Aware”. Most current devices teach us to do something – push a key or make a particular motion – versus having it learn what we want it to do based upon our current context. For example, when you walk into a meeting your computer should automatically load the correct presentation based upon knowing you are in the meeting by sensing your location and knowing the agenda based upon your appointment book. At the same time your phone should switch to vibrate mode or only interrupt with certain calls.
- Syride – Their product will reduce unsubstantiated bragging in extreme sports starting with surfing and paragliding by providing actual measurements. Inside their unit is a multitude of sensors – three axes accelerometer and three axes magnetometer plus a temperature sensor – along with a global positioning unit (GPS) that tracks a users performance during their fun. The user can then upload the data to their online community to share results and to help improve their performance according to Romain Lazerand (Business Development Consultant).
- VectorNav Technologies – Jeremy Davis (Director of Engineering) described how they start with consumer grade MEMS sensors and through combining the data from multiple types of sensors (using improved algorithms and device specific calibration) they are able to significantly improve the accuracy and performance of the devices. This allows them to supply sensors for higher performance applications such as industrial and navigation at a lower cost. In addition, by merging the sensor data they can filter out (ignore) spurious data. For example, in their demonstration he showed how the presence of a strong magnet did not influence their directional heading reading even though they use magnetometers.
- Xsens – By combining the data from eighteen MEMS sensors worn on the body and processing the data in realtime their Xsens MVN product is able to perform motion capture anywhere without cameras. This can be used for numerous applications from movie making to medical research according to Casper Peeters (Chief Executive Officer). He also provided a preview of a consumer grade system that will be available at a lower price point that could be used for gaming and exercise coaching.
Though each gave a good overview in their presentation, more impressive was to actually try the products in the exhibit hall. Obviously, the ones that were complete products gave a better showing but the the technology demos were also quite convincing.
Analysis and Summary
Is a trillion dollar and / or unit market feasible? Below is an approximate plot of my impression of each speaker’s answer to that question. Many were intentionally vague and the answer is likely to change drastically depending on the average selling price per sensor (shown for some of the answers provided as ranges). What is clear from all the answers and the empty upper left quadrant is that the overall market needs to have an average selling price of less than $1 per sensor even at lower volumes.
In summary, here are the key takeaways from this informative event:
- The value is in higher order solutions not in producing the MEMS sensor. These solutions include Sensor Fusion, software, and hardware (micro controllers, subsystems, IP cores, etc.) to provide meaningful and actionable data to the rest of the system not just raw measurements.
- The rate of adoption is severely influenced by price. To accelerate adoption and increase volume, prices need to continue to decline.
- Coordinated coopetition is needed via an organized roadmap to both reduce costs and to solve the technical challenges of simulation, packaging, and test.
- To spur additional applications, vendors will need to educate software developers and product designers as to the capabilities and how to effectively implement solutions.
- Don’t always believe market estimates in new markets with new applications as they can be widely overoptimistic or pessimistic.
It was a pleasure meeting everyone at this event and I look forward to continuing the discussion. Please share your thoughts or questions in the comments below.