With a saturated smartphone market where the number of sensor units per device appears to have peaked, MEMS product companies are pursuing new market opportunities. Two emerging sensor-rich applications are drones and virtual reality (VR). Drones use an inertial measurement unit (IMU) for positioning, orientation, maintaining direction and flight paths, and a slew of other sensors for stability control (tilt sensor), navigation (magnetic sensor) and elevation control (pressure sensor). VR headsets and accessories rely on fast and accurate motion sensors to track movements and gestures. So how are the major consumer MEMS product companies approaching these new opportunities? We recently sat down with Behrooz Abdi, CEO of InvenSense, at the 5th annual InvenSense Developers Conference in Santa Clara, California to get some answers about the latest trends in MEMS and sensors.
MEMS Journal: How is InvenSense doing these days? What new projects are you working on? What are the main milestones on the agenda for the rest of this year and 2017?
BA: We are very busy, but good busy. There is a lot going on in mobile. There are approximately 2.2 billion phones produced and sold each year, and about half are considered smartphones (i.e. they have downloadable apps and a screen). Of the 1.1 billion smartphones, about half are considered high-end, which we define as gyro-enabled to accommodate complex applications like image stabilization, advanced navigation, and contextual awareness. Using a total available market (TAM) of 550 million high-end smartphones, we have a 50% market share. The focus has up to now been in the flagship market segment. We are currently driving adoption into mid-tier smartphones for applications like augmented reality (AR) and navigation. Pokémon Go is speeding up the market traction for gyros and mid-tier smartphones now need 6-axis for product differentiation. This increases our TAM in mobile by 400 million to 500 million units per year. Other markets we are involved in are drones and wearables. The wearables market is fragmented with, for example, smartwatches and fitness trackers. FitBit has become a leader and is large, but there are many smaller players. InvenSense has developed a platform for wearables -- the sensor hub Firefly with integrated motion sensors. VR is becoming interesting and we have a platform for that too. We are also looking at high-end audio applications for our microphone technology -- for example smart speakers such as Google Home.
Over the next year or so, InvenSense has milestones in mobile for all-in-one product solutions for user interface (UI) and optical image stabilization (OIS) using a 6-axis device. We are working on high-end microphones and a time division multiplexed (TDM) microphone array with up to 16 devices. The ultrasonic fingerprint sensor is coming and we have taped out our first chip. We are also making progress in automotive and have a system integration partner (Panasonic) for developing 6-axis devices for safety applications.
MEMS Journal: What is your long term vision? Where will InvenSense be in five years?
BA: In five years we are aiming to be a pure-play sensor company with a broad-based range of products in multiple industries. We want to provide end-to-end solutions in markets ranging from mobile and the IoT, to industrial and automotive. In addition, we want to grow in areas other than mobile. One reason is that mobile reached saturation this year. The overall goal is to grow the revenue by 15-20% per year.
MEMS Journal: What are the main industry trends impacting InvenSense? What are your main challenges right now?
BA: This is a challenging year for mobile. There is price erosion due to overcapacity of MEMS devices in the market. Industrial and automotive markets are moving towards sensor integration. The automotive market is still mostly based on single-axis motion sensors, but it is trending towards the integration of multiple axes for motion sensors. We have a path to a 7-axis device specifically for drones. This is a product with 3-axis accelerometer, a 3-axis gyroscope, and a barometric pressure sensor. Location-based services are becoming important, as well as asset tracking, and retailers wanting to track consumers.
MEMS Journal: Consumer electronics OEMs are increasingly investing in sensor technology. Is there a risk that OEMs will start making their own sensors?
BA: No. Many companies can build and sample MEMS, but making millions is difficult. There is a very large barrier to entry.
MEMS Journal: Can you talk a bit about other emerging markets such as home automation, augmented reality, and IoT devices? What role is InvenSense playing in these markets? Which of these markets is the most promising for your business and which of these segments do you have the most traction?
BA: VR is very promising along with drones. This is the start of a revolution for VR that goes beyond just gadgets for early adopters. Over the next 2-3 years it will be mainstream. There are some really good head-mount displays coming out this year and a lot of content is being developed. As the cost structure comes down, there will be more adoption. In addition to the headsets, there are also the controls and accessories. Head-mount displays are very sensor rich, and will include audio and user authentication for in-game purchases. These applications require motion sensors with low latency, low noise, and high performance. Drones are mostly for hobbyists right now, but they are becoming mainstream. One example is drones for in-building monitoring, like a dynamic security system. Robots are currently used in high-end applications, but becoming personalized. There are service robots in hotels and Softbank is introducing robots in their stores. For home automation, InvenSense is targeting smart remote controls with gesture control and smart speakers for our high-end microphones. We are also looking at bringing environmental sensors into the home. The goal is to create sensor rich platforms for the next IoT wave, which is trending towards location and contextual information detection using sensor clusters and chipsets.
MEMS Journal: InvenSense has added a lot of software resources over the last few years and developed complete software solutions such as Coursa Retail, a location-based, in-store shopper analytics platform. How is the traditional MEMS component business changing and how are you capitalizing on software? How has Coursa Retail been received in the market?
BA: We now have between 80 and 100 software programmers. We really need to understand what customers are doing with sensor data to adapt our products to customer needs. From there we can identify what imperfections in the MEMS devices are impacting the various use-cases and what can be done in software. It really starts at the basic level of the MEMS sensor. InvenSense is moving to turnkey solutions with embedded software. We want to develop and provide full sensor solutions. For example, when we entered the Chinese mobile market, we used sensors from other companies and our own algorithms to demonstrate image stabilization functionality. Then we compared this to our own turnkey solution, which included our sensors, to show why a better gyro was needed for this application. This helped us create a stickier product. For the longer term, we are looking into how we can monetize sensor data in the cloud. However, we have a different strategy for selling to customers our cloud based service. For example, with Coursa Retail we go directly to retailers and offer a software solution. The customer only needs a beacon in the store and the data is coming from the customer’s smartphone. Since very little infrastructure is needed, it creates a low-barrier to enter this market. InvenSense is trying to remove the friction of implementing infrastructure by leveraging, for example, wireless beacons.
MEMS Journal: What’s your IoT strategy? Many industry analysts are saying that most of the value in IoT applications will be captured by data and analytics companies, rather than sensor providers. What are your comments on this?
BA: InvenSense has a two-part engagement model in the IoT space. First of all, we have a sensor hub (Firefly) and we are also in parallel rolling out an updated version of Sensor Studio, which was announced at IDC 2016 on October 20th. Sensor Studio is a development platform that helps developers create sensor algorithms and run them on pretty much anything that has an ARM core. This is a very similar approach to what has been done for microcontrollers for years, but for sensors. We are aligning with distribution partners like Avnet and Digi-Key, and equipping them with tools that remove the barriers for customer adoption. We are trying to take math and size out of wearables, providing both the sensor algorithms as well as the small footprint 6-axis motion sensors with a sensor hub. The Firefly sensor hub also allows the use of third-party sensors, including environmental sensors. The second engagement model is to develop software location and activity software solutions that combine embedded software and cloud processing. The intention is not to compete with the large analytics companies, but to bring valuable features to these providers. One example is better location resolution for ride-share companies and detecting what type of activity the passenger is performing.
MEMS Journal: What are the most interesting trends for sensor hubs, with or without integrated sensors?
BA: For us, the multi-sensory aspect of sensor hubs is the most interesting. This includes the combination of sensors and our Sensor Studio for third-party sensors. The differentiation for sensors hubs is to have both sensors and algorithms. We want to do what Moore’s Law did for digital and similar to how application processors (AP) have become integrated solutions with multiple functions. InvenSense has a CMOS/MEMS platform that scales really well. There will be 10-axis IMU devices in the future -- self-contained, all-in-one sensors.
MEMS Journal: Are you seeing an increased demand for combo-sensors? What are the most popular combos and what are the applications?
BA: OEMs do not necessarily demand combo sensors. Until we show up and tell them what a combo sensor can do, they will not request it. For example, InvenSense brought a 7-axis solution to drone companies. Before that, they did not realize that this could be done. It is really up to the sensor companies to drive the development of new combo sensors. In the earlier days of InvenSense, mobile customers did not demand a 6-axis device, but when InvenSense developed a combo sensor it made sense to the OEMs.
MEMS Journal: What are your thoughts around the rapidly declining MEMS unit prices? Is this the main reason why your revenues have been growing, but profits have been flat?
BA: There are two sides to this. The usual situation when you only have three competitors in a market, everyone makes money. This is not currently the case in mobile. It is a special situation. The two other competitors (i.e. Bosch and ST Microelectronics) have developed a massive overcapacity. This is heavily impacting the unit prices. I am not worried about the unit prices long-term. Gross margin has suffered in the past, but we feel that we have enough cost reduction in the pipeline. We are also adding features for new adoption markets that do not have the cost structure of the mobile market. On the other hand, new applications are taking advantage of the low-cost consumer sensors. VR would not be so easy without it, especially considering that headsets must come down to a few hundred dollars for mass-adoption. Sometimes we need to take risks in order to innovate.
MEMS Journal: How do you see mobile and consumer electronics markets adopting new sensors? For example, iPhone 7 does not include any new sensors or sensing capabilities. Is this an indicator that the number and type of sensors in smartphones have peaked?
BA: It comes down to use cases and it will go through cycles. A few years ago, a large Asia-based smartphone OEM adopted humidity and temperature sensing, but no use cases were found so they removed them. Biometrics and new user applications could drive new sensors, for example OIS for dual-cameras now in flagship smartphones from Apple and Huawei. Another example is the adoption of pressure sensors mostly in flagship smartphones and not in lower tiers. There might be external factors playing a role as well. For years, regulators and the FCC have been working on a first responder’s mandate that would require a barometric sensor in phones to make it easier for emergency personnel to locate people in distress. Biologic sensors for scanning and determining the quality of produce is another example. Some trends are gimmicky, like some of the gesture recognition in smartphones. But if you can do real things, it has value. We are very actively looking at new applications.
MEMS Journal: What are your thoughts on alternative technologies and materials for new generations of smaller, lower-power, more-integrated, and lower-cost sensor devices? Do you see a shift to “package-less” devices and WLCSP?
BA: Yes, I think there is a shift towards package-less solutions. That is where we are driving. InvenSense’s CMOS/MEMS technology is a good fit for that. The trick is to take the mystery out of MEMS. We have solved many issues with MEMS to get to where we are now and to make it look more like CMOS. We are still trending towards this goal. There are no SPICE models for MEMS and many sensor simulation tools are custom. Because of this, we have, developed our own proprietary in-house MEMS development tools. MEMS is currently similar to where RF was in the earlier days. No one thought RF would become compatible with CMOS processing. The same can happen with MEMS.
MEMS Journal: What’s been going on with your MEMS microphone business? Can you share some specifics on the number of mics shipped?
BA: We have been addressing high-performance markets. We have not made much progress in mobile for lower-end applications. Instead, we have gone after non-mobile markets and have good traction. We have many design wins for smart speakers and high-end headsets. The percent of revenue from microphones within our total business is still in the single digits. Next year this should go to double digits. We would be relying on mobile for this growth, with microphone arrays and higher-end products. The lower-end microphone market is highly competitive where the value is in minimizing package costs. With higher-end products, the value will be moving from packaging to the SoC. We are targeting applications that need 67dB SNR and have devices with 70db and higher. Our time division multiplier (TDM) version can include up to 16 units for beam-forming applications.
MEMS Journal: How much business do you currently have with automotive applications? Is there a specific application where you have the most traction? What are your plans for automotive going forward?
BA: In automotive we are targeting two segments -- in-cabin and safety. For in-cabin we have a product and navigation software. We are making good progress with the system integrators. We are not shipping in volumes yet, but are currently in the design-win phase with a 6-axis device leveraging our mobile solutions. In the safety segment, we have a recently announced partnership with Panasonic. This will allow us to penetrate into the safety market with an integrated 6-axis device.
MEMS Journal: Your company slogan is “sensing everything”. Are you really planning to develop all types of sensors? Will this cause a problem for your company focus?
BA: We want to be a pure-play sensor solution company and may eventually do all types of sensors. We are, however, doing it in phases and in a scalable manner. We went from motion sensors and developed a membrane technology for microphones and pressure sensors. Then we brought in the microphone division from Analog Devices, which had a different technology from what we were developing internally. This allowed us to develop fully-integrated SoCs. Then we went into ultrasonic sensing where we identified needs for differentiated fingerprint sensing solutions. We are mainly looking into applications and solutions where our existing technology can apply. We have, for example, not done anything with RF and timing devices. We have chosen to focus on sensors even though our technology can apply in other areas. We also have to watch our R&D spending.
MEMS Journal: What has your experience been with business in China? What’s the main challenge of doing business there? What do you think about the Chinese government increasingly funding their domestic sensor suppliers and how will this trend impact InvenSense?
BA: There are great opportunities in China, especially in the IoT. Mobile has been good and the drone market is good. Even with funding, the Chinese companies lack the fundamental technology. We do not see any Chinese MEMS companies coming up over the next 5 years to compete. They might eventually, so we will keep innovating to maintain our leading position. Overall, China is a more of an opportunity than a challenge.
MEMS Journal: What are 2-3 most interested MEMS and sensors startups which you’ve seen recently? What’s interesting about these companies?
BA: I cannot mention companies specifically, but technologies in environmental sensing are interesting. There are a few companies looking at this. Other technologies we follow are sensors for medical applications, microfluidics, and actuators. Energy harvesting is interesting for self-powered sensors. Medical sensors are interesting, but the time to market is long and requires a lot of patience. We are currently set up to pursue applications that have a faster track to the market.
MEMS Journal: InvenSense has made several acquisitions in the past. Is this an ongoing effort? What specifically are you looking for -- new sensor technologies, software algorithms, or other?
BA: The primary path is growing organically, but we are looking at acquisitions to speed up time to market. This was the case with microphones. We did not go after Analog Devices for their microphone business unit, but grabbed the opportunity when it became available. If there are technology or business gaps to fill, we will take action. At this point we are well-known within the ecosystem. We usually get a first look at new companies and technologies. That is a privilege.
MEMS Journal: What are the main limitations for improving performance in MEMS?
BA: There are physical limitations in terms of stability and drift over time. This is a predominant factor in all sensors. The InvenSense technology platform represents a fundamental improvement in sensor technology. Now we are discovering other ways to overcome additional limitations. There are fundamental challenges to make MEMS look more like CMOS, for example, to eliminate drift over time and potentially temperature effects. There is a lot of work still to be done, but the market will demand it. At times we hit against “good enough”, but then other applications come up with new requirements. For example, when we had 3G networks for mobile phones, all we wanted to do was talk. Then the iPhone came along and changed everything.
*********************************************
This article is a part of MEMS Journal's ongoing market research project in the area of MEMS and sensors for consumer electronics applications. If you would like to receive our comprehensive market research report on this topic, please contact Dr. Mike Pinelis at mike@memsjournal.com for more information about rates and report contents.
Copyright 2016 MEMS Journal, Inc.