By Sara Verbruggen
Initiated by Apple’s launch of the iPhone, the subsequent explosive growth of the smartphone market has provided the MEMS industry with one of its biggest opportunities to supply high-volume demand. But if motion sensing in our portable electronics – enabled by accelerometer and gyroscope MEMS applications for example – is the tip of the iceberg for MEMS technology how can the semiconductor industry ensure that high volume markets like consumer electronics benefit from all that MEMS potentially has to offer.
As the MEMS industry evolves, in terms of further diversification of device applications in higher volumes, this creates manufacturing challenges.
‘Organizations like MIG are helping to set standards across classes of devices in terms of specifications, rating, test interfaces, and system interfaces, and this is a great advancement in helping the industry to grow. On the manufacturing side though it is unlikely that a “standard” MEMS flow will emerge even within individual foundries except for very specific and limited types of MEMS – Invensense NF Process is an example of an attempt at this,’ comments Silex Microsystems’ VP of marketing and strategic alliances Peter Himes.
The emergence of MEMS technology over the last decade into high volume markets – consumer electronics especially – has presented the semiconductor industry with the challenge of designing and fabricating devices with different functionalities (as opposed to focusing on scaling down while ramping performance). This has paved the way for electronics in industries as diverse as healthcare, energy, security and environment. The long-term growth of MEMS depends on functional diversification but also being able to manufacture devices for these various applications in significant volumes and bringing down cost.
More than Moore techniques and processes
Wafer-scaling fabrication and process technologies, to enable these ‘More than Moore architectures’ are beginning to become established in MEMS manufacturing, for high volume markets.
SEMI’s chief marketing officer Tom Morrow says: ‘To be competitive in high-volume MEMs markets, 8” production equipment and economies will be, if not already, needed. Deep reactive ion etch (DRIE) “tuned” for MEMs technologies are also required, coupled with advanced cleaning solutions such as plasma. Bonding is, with DRIE, the other key MEMS-specific technology, used for wafer level capping and wafer level packaging.’ Critical concerns include providing good hermetic solutions to maintain performance of sensitive moving parts like gyros, while taking up less area on the wafer with bond lines. ‘The bonding process tends to take time, so throughput is typically low. Room temperature bonding and temporary bonding are areas of major improvement,’ adds Morrow.
DRIE and wafer bonding are the technologies subject to significant process improvement as both technologies are increasingly used in the mainstream semiconductor industry for 3D-TSV. In addition packaging and bonding technologies today support increasing standardization.
‘While contact and proximity aligners remain prominent lithography tools for MEMs, there is some movement towards projection steppers for better CD uniformity and automated 8” volume production,’ according to Morrow. Tools also need to be able to handle thin wafers and manufacturers also demand better overlay precision.
TSV is a critical technology, agrees Silex Microsystems’ Peter Himes. The company has specialised in TSV integration into MEMS since 2005 when its Sil-Via technology went into first production. This process, developed for the mobile industry, consisted of an all-silicon interposer for 2.5D integration of a MEMS microphone and ASIC onto a silicon substrate which was then solder- bumped and mounted directly onto the PCB.
‘Since then, we have been developing more TSV options for our customers, including TSV for buried cavity MEMS, TSV for capping solutions of either MEMS or CMOS, and both metal TSV and TGV through glass substrates for RF and power applications,’ says Himes.
As MEMS companies increasingly move beyond competing on manufacturing technology to competing on functionality, more of TSV/WLP packaging solutions will become widely-used platforms, predicts Yole Développement. This would also make more use of the outsourced infrastructure to reduce costs and speed-up development time.
‘Today, a few MEMS companies such as VTI, STMicroelectronics, Robert Bosch or MEMSIC have successfully implemented 3D wafer-level packaging concepts by using TSV/TGV vertical feedthrough, redistribution layers, and bumping processes to directly connect the silicon part of the MEMS/sensor to the final motherboard but without using a ceramic, leadframe, or plastic package. We believe this trend will be accelerated even further with the shift to 200mm wafer manufacturing for MEMS: it just makes sense to use wafer-level packaging, because as soon as you can add more dies on a wafer, it is more cost-effective,’ says Eric Mounier from Yole.
AMAT’s Mike Rosa points out that wafer-scale integration techniques, to enable more device functionality on a per die area basis, in combination with system-on-chip technologies to enable greater intelligence on die is becoming a standard requirement for more advanced MEMS. ‘The end-users (system integrators – like Apple or Samsung for example) now require the MEMS device to do a lot more of the signal processing than has traditionally been the case – hence MEMS designers have to include more signal processing (CMOS) capability on die,’ says Rosa.
The fabless approach in the MEMS industry is now well-established, where, in order to speed up MEMS development device cycles, foundry companies partner with designers to provide them with process modules around which designers can develop MEMS devices.
But for the fabless model to facilitate the development of more differentiated and disruptive MEMS and to ensure companies remain competitive manufacturers need to be able to embrace and adopt new manufacturing processes and material technologies – which accompany disruptive new MEMS devices. ‘In the foundry space, it’s the foundry partner who is strongest in technology development that will win market share – this there is already a clear ‘pecking order’ with the big three foundries today and that is for a very good reason,’ says Rosa.
Silex is an example of a successful business servicing the fabless segment, through its program with AMFitzgerald. ‘The fact is that new companies cannot afford the cost of building a MEMS manufacturing line, and need a foundry infrastructure to get their products to market,’ says Himes.
Several key factors point to a strengthening fabless market in the long term, he observes. These include an ongoing reduction in overall development times for MEMS over the past two decades, lowering the time to market for new MEMS devices ‘though Yole is correct in saying that it needs to come down further,’ he adds. Increasingly fabless start-ups are driving innovation in MEMS-based functionality. ‘The percentage of MEMS revenues which comes from components not on the market before 2006 has been steadily growing, pointing to increased diversity and expansion of the MEMS- enabled market,’ says Himes pointing to a recent iSuppli presentation.
‘In terms of what works, Silex’s systematic SmartBlock-based approach toward process integration coupled with our defined new product introduction (NPI) process has proven to be the best way for us to manage the risk and uncertainty which comes with any process development. While customers always want shorter time to full production, an early focus of our customer programs is to get the customer fully functional samples as early as possible so that the rest of the component or system can be developed,’ Himes explains.
According to Mounier a successful fabless model relies on a MEMS designer, or similar business, finding a reliable foundry working on the long term. ‘Depending on the application, the foundry will have to be competitive on cost (consumer, automotive) or performances (defense, industrial applications). However, as many new MEMS devices are emerging in for new applications, such as touchscreens and flat speakers, MEMS foundries must be able to think about adapting the customer design to their own process flow.’
The RocketMEMS program run by AMFitzgerald & Associates is a good example. The company has defined a product design platform for rapidly commercializing semi-custom MEMS devices (pressure sensors is the first area) based on a pre-qualified manufacturing flow at Silex. ‘We think that this is an efficient path toward design enablement that can avoid the “one product, one process” paradigm in the long term,’ says Himes. Customers would be prioritizing time to market and customized form-fit-function over fully customized and optimized MEMS process flow. ‘We can envision many more such programs being set up worldwide, and thereby expanding the capability of doing MEMS design from the PhD level down to a broader class of component design engineers,’ he adds.
There are various challenges in the MEMs industry, owing to both the required process craftsmanship seen in advanced devices and the sheer proliferation of device types. Morrow observes: ‘Foundries continue to address these challenges through process capability improvement, and are benefitting from a maturing design process ecosystem that understands the need for integration with manufacturing, particularly in high-volume segments such as inertial sensors, microphones, and optical MEMs. Lower volume products, highly specialized device types, unique packaging or ASIC integration requirements seem to support IDM-type manufacturing.’