MEMS moves into the mainstream
4 mins read
Like the carbon nanotubes featured in this issue's Cover Story, MEMS is often regarded as a technology of the future.
But that's not the case; MEMS devices have been with us for more than a decade, while research and development activities started a further decade back. However, the advent of approaches like the Internet of Things are set to propel MEMS to the forefront of technology and a number of companies are looking to take advantage of the opportunities.
Jay Esfandyari, MEMS product marketing manager with STMicroelectronics, said there has been 'significant progress' since the first MEMS devices appeared early in the 2000s in mobile phones and in games consoles.
He pointed out that MEMS was known as microsystems technology in its early days. The phrase 'micro electromechanical systems' – coined by a US academic – has become the standard way of referring to devices with at least one dimension of less than 100µm. "They also need to have at least one moving part," he continued.
Despite consumer electronics companies being amongst the early adopters, the first applications for MEMS devices came in the automotive sector – particularly airbag actuation – and this remains a significant market.
"The first successes came with accelerometers," Esfandyari noted, "but it then took five or six years to do gyroscopes successfully." Now, more complex devices are found in a wide range of products and industries.
According to Esfandyari, three parameters drive the adoption of MEMS devices: integration; cost; and power consumption. "How small can we make them, how much can we get into the package and how much power does it need."
Because MEMS devices, unlike their distant cmos cousins, feature moving parts, they need to be manufactured using special processes. Despite two decades of research and development, MEMS devices have yet to be integrated with digital logic on the same die.
"The mechanical structures have to be manufactured on a special process," Esfandyari explained. "It was adapted from a standard cmos process, but there is more etching than with digital logic, as well as more polysilicon and so on. There are similarities, but MEMS and logic won't be on the die any time soon. If you try to combine the two, you will lose flexibility."
ST's MEMS products are manufactured on a proprietary process called Thelma – Thick Epitaxial Layer for Microgyroscopes and Accelerometers. The 0.8µm surface micromachining process, which runs on 200mm wafers, combines variably thick and thin polysilicon layers for structures and interconnections. This is said to enable the integration of linear and angular mechanical elements on one die, delivering cost and size benefits.
Because the MEMS element moves, changes in capacitance can be detected. This information is passed to an asic, which removes noise and calibrates the data, providing what Esfandyari called a 'clean output'.
"Each supplier has its own MEMS process," he continued. "There is no industry standard process, although there have been discussions about this for years. Proprietary processes mean suppliers can differentiate themselves."
Esfandyari said suppliers are now under more pressure to develop smaller parts. "Customers expect it," he said. "We've had to integrate further and one way we've done this is by bringing mechanical structures onto the same die. There is also one asic per package: it's a system in package approach."
The LSM303C, one of ST's latest MEMS products, is a compass module with six degrees of freedom – a three axis accelerometer and a three axis magnetometer. It comes in a 2 x 2mm package with temperature sensors and a fifo memory. The accelerometer can measure ±8g, while the magnetometer measures 16Gauss.
While MEMS production has been the province of specialists, foundries are taking a much closer look at the market than before. The reason? Predictions suggest the market might be worth more than $12billion a year by 2017, with demand driven by wider use of MEMS in consumer electronics, as well as in industrial applications and in the Internet of Things.
Gregg Bartlett, chief technology officer of Globalfoundries, provided a keynote address to the recent MEMS Industry Congress. He said a 'big window' is opening for fabless – even fab lite – companies to move into large scale manufacturing of MEMS devices. In particular, he pointed to foundries providing such advantages as modular technology platforms, standardised interfaces and larger wafer sizes for economy of scale. Foundries, he contended, can address design cycles that are months, rather than years, in length. When it comes to MEMS, he said, foundries are moving from being 'contract manufacturers' to becoming 'solutions providers'.
"As the MEMS industry's impressive growth continues," he said, "we must work collectively to move beyond the specialised tools and processes, following a 'one product, one process' approach."
The foundry claims to have its sights set on becoming a leader in high volume MEMS manufacturing by 2015 and is pursuing an 'aggressive strategy' to achieve that. It claims it has already made 'major design wins' with tier 1 customers.
Rakesh Kumar, senior director of Globalfoundries' MEMS programme, has said: "CMOS foundries need to adapt by developing new standardised process capabilities, technology platforms and IP to enable high volume manufacturing of MEMS devices. They also need to support advanced packaging technologies, such as system in package and chip stacking with through silicon vias, to allow MEMS device designers to pack more functionality into one package. Our success in embracing this new model depends on our ability to offer solutions across the value chain – from design support and process technology to assembly and test. Close partnerships within the MEMS ecosystem, and customer collaboration, are equally vital."
Globalfoundries' MEMS capacity is centred on Fab 3E in Singapore, where it has recently entered an agreement with the A*Star Institute of Microelectronics, as well as with the Abu Dhabi based Masdar Institute of Science and Technology, to 'develop and advance' MEMS technologies for automotive, aerospace, consumer, healthcare, industrial and mobility applications. Amongst the devices to be developed are inertial sensors, energy harvesters, nano optomechanical sensors and ultrasonic transducers. Globalfoundries says the deal will enable it to expand its IP portfolio to address high volume MEMS manufacturing.
Kumar said foundries are capable of using their processes to enable new product development, to provide a fast ramp to production and competitive manufacturing costs. "We envision taking the role of a single supplier to provide the complete manufacturing solution that will allow our customers to focus on product design, firmware, applications and system level support. We can achieve this by not only offering device fabrication services, but also extending it to complete back end solutions."
Asked about the Internet of Things, Esfandyari said it cannot exist without sensors. "The only way to make things smart is through the use of sensors," he concluded. And it is his belief that the billions of systems which will be hooked up to the IoT will need billions of MEMS devices.