
Fabless inertial sensor startup mCube Inc. is just about to enter phase three of a startup’s progress. That’s why it has just secured $37 million of Series C venture capital, CEO Ben Lee told eeNews Europe.
Lee, who joined to lead the company in January 2013 as CEO, explains that in a startup’s first phase it has a little bit of money for development work and to create a product, but no customers. In the second phase it has a product but can only support – and only listen to – a few customers. The third phase – for those that make it that far – is the growth phase where a company starts to have revenue and can expand its product and customer bases.
But in that third phase it remains crucially important to keep making the right choices in terms of customers and application areas to which a company is applied.
mCube (San Jose, Calif.) is going to use its $37 million Series C money to maintain focus on the entry- to mid-level smartphones in the Chinese market while pursuing innovative applications in wearable equipment and the Internet of Things, where it feels its combination MEMS sensors have an inherent advantage.
The company was founded in September 2009 with the mission of commercializing a method for integrating MEMS motion sensors above electronic circuitry in a standard CMOS wafer fab using through-silicon via connections. The process includes hermetic sealing of the assembly.
The inherent advantage that Lee talks about is primarily size. The use of wafer bonding and through-silicon-vias (TSVs) allows mCube to produce a combination inertial sensor in a package measuring 3mm by 3mm, without needing bond wiring and bond pads inside the package.
Three generational approaches to making inertial MEMS. Source: mCube.
While market leaders Bosch and STMicroelectronics use traditional hybrid manufacturing techniques they have been challenged more recently by InvenSense with its stacked chip approach, Lee indicated in a slide presentation. But all three will now be challenged by mCube with is monolithic CMOS approach which can dispense with bond wire and bond pads to produce smaller size and lower parasitic capacitance resulting in better accuracy, he claimed.
The approach gives mCube an advantage in terms of die size, packaged footprint, sensor accuracy and energy consumption, Lee claims. This in turn will not only let mCube steal design slots from its better-established peers in smartphones and tablet computers but will also help the company help drive wearable equipment and Internet of Things markets.
Small is beautiful
mCube has spent four years in the first two phases of the startup’s progress but is now easing itself into the third phase. The company has already shipped more than 60 million units into a range of smartphone, gaming and tablet reference designs since 2012 and its MEMS sensors are featured on the approved vendor lists of chipset partner – and investor – MediaTek.
"We are primarily in tablets and smartphones for the Chinese market, and some Bluetooth headsets and smartwatches. We’ve focused on the Chinese market and now have 60 to 80 active customers," said Lee.
But what about high-volume, high-profile design wins in Apple and Samsung equipment that are surely the goal of every consumer electronics oriented component supplier?
Lee explains why such a design win is not always good news. "Apple and Samsung tends to be a two-year design cycle. It tends to be hit or miss. And when you kicked out it is very disruptive to the company," he said pointing at companies that have been more or less destroyed by the loss of an Apple design win. The Chinese market is faster moving and much finer-grained with scores of "white-box" companies designing and making equipment for brands in China, said Lee.
"There are a slew of ODMs [original device manufacturers] supplying the brands in China. Part is brand supply then there is also a domestic no-brand market. And then there are also emerging markets in India, Africa, Indonesia and elsewhere. There is a market of about 600 million phones shipped annually served by 100 ODMs in China that go into local and international markets," said Lee. "But you have to behave like a Chinese supplier. It’s the wild, wild East," he said. But it offers a number of things that are much better for a startup than an Apple or Samsung design win. "It’s fast time to market, fast time to money and fast time to feedback," Lee said.
"And now there is wearables. mCube has the world’s smallest accelerometer – that’s confirmed by Yole Developpement," said Lee. "So our timing is very good."
Lee explained that while the motion of gaming and phone handsets and tablet computers is definite and relatively easy to detect, wearables are throwing up the need to detect subtle movements. "Trying to detect very slight movements in wearables is pushing companies to deploy three or four sensors as far apart as possible on the PCB, which means the sensors must be small. So it’s not a 1:1 attach rate but multiple separate devices."
Fabless with TSMC
But mCube is fabless and so the manufacturing of its MEMS sensors comes courtesy of foundry supplier Taiwan Semiconductor Manufacturing Co. Ltd. What’s to stop TSMC offering the process to others creating competitors?
"We co-developed the process with TSMC," said Lee. mCube has also filed more than 100 patent applications. Lee does not rule out the possibility that other companies might use manufacturing services from TSMC of a similar nature but added: "It’s true we are the designers but in MEMS design and manufacturing are highly integrated."
Bonding wafers and use of TSVs produces a monolithic CMOS sensor just 600-micron thick. Source: mCube.
Lee points out that one of their key patents covers how to accurately control the depth of etching for a TSV so that there is a connection but no damage on the primary wafer that carries the electronic circuitry. Without that technique the approach is closed off to potential rivals. With it mCube is able to create 14-bit resolution sensors when other companies struggle to achieve 8-bit resolution, Lee said.
Grain of sand
With sensors in 3mm and 2mm packages mCube is already down at the same sort of size as a grain of sand – coarse sand comes in at 1mm to 2mm.
But what if, with the monolithic construction mCube could forego the plastic packaging and directly bond its hermetically sealed die to the PCB? "The z-height goes even lower," said Lee. "We have to look at structural integrity issues and how the cavity inside the sensor performs. But we have alpha customers and it’s very exciting," said Lee.
Lee also has an mCube take on the sensor fusion, sensor hub debate. Does it mean that mCube will be adding microcontroller circuitry for software programmability within the company’s tiny combination sensors?
Lee said there are three main approaches:
1) running the sensor fusion software on the application processor, such as a MediaTek octocore;
2) having a separate sensor hub IC, such as Apple’s M7;
3) and putting an MCU inside the MEMS package.
The decision is usually based on the lowest power for an always-on solution so that motion of the equipment can be used to wake up the circuitry. "Most of our customers don’t want us to do it in the third option. But that doesn’t mean they don’t want us to write the software. They want us to provide a library of algorithms for accurate pedometer and so on." As an extension of this mCube has written a couple of games that can show of the ability of the company’s iGyro component.
For now Lee is determined that mCube should follow the business model of its investor and partner MediaTek. "China is our first focus. We’re not concentrating on any other customers but those. But our reputation for smallest size does mean we are starting to get cold calls for the wearables market," said Lee.
Related links and articles:
News articles:
mCube raises $37 million for motion sensor push
mCube launches 9-DoF inertial sensor
Yole: MEMS in mobile market on 14% CAGR
Will sensor fusion drive neuromorphic computing?
Internet of Things: Making the component industry think inside the box
