By Sara Verbruggen
Semiconductor Design & Manufacturing discussed ‘More than Moore‘ (MtM) standardization topics and challenges with Peter Himes, VP of marketing and strategic alliances at Silex Microsystems, Dr Eric Mounier, senior analyst, MEMS devices and technologies at Yole Développement, Tom Morrow, chief marketing officer at SEMI and Mike Rosa, senior global product strategic marketing manager – emerging technologies, 200mm components and systems group at Applied Materials. What follows are excerpts of that conversation.
SemiMD: What is the scope and potential for further standardization in MEMS fabrication – for example, consumer markets such as handheld/portable electronics demand MEMS that are smaller and lower in cost, but to what extent does the diversity in MEMS devices in terms of their functionality, applications and manufacturing defy standardization and how can this be overcome?
Peter Himes: In the MEMS industry I would say that the opportunity for standardization today is limited, but the potential for more standardization in certain areas is sure to develop. Think about the mechanical goods and products you use each day. Are all corkscrews identical? Do all blenders look the same? Are they manufactured exactly the same? Yet they all perform the same function. This is where MEMS is today, with differentiation through MEMS design. You can say that for all blenders the motor element can be the same and products can differentiate with functionality and product design, and maybe that is a model for certain classes of MEMS moving forward.
Eric Mounier: Standardization is definitively an issue for MEMS manufacturing. However, standardization can occur at two levels; packaging standardization and front-end process standardization. Both are underway in the MEMS industry.
Referring to packaging standardization, MEMS types of packaging are more complex than most standard IC packages because they require a System-in-Package type of assembly. Additionally, sensor packages are generally quite bulky and can have very specific constraints like a module with a cavity, a hole in the substrate or metal lead for pressure sensors and microphones, an optical window for optical MEMS, or a full vacuum hermeticity at the die level. As a result, standardization is becoming increasingly critical to support the massive volume grow in unit shipments along with decreasing overall costs associated with MEMS and sensor content, in particular related to their packaging. If we look at one specific MEMS example, a microphone for instance, these are all packaged the same way: BGA/LGA laminate PCB substrate + SiP module assembly with wire bonding + metal lid or PCB cap with integrated shielding + “hole” for air access.
Tom Morrow: The history and dynamics of the MEMs industry has not facilitated MEMs manufacturing standardization as we have seen in the semiconductor industry. This is partly due to the diversity of devices where rapid adoption of some early products — such as tire pressure monitoring, air bag deployment sensors, ink jet printing heads, and others — initially encouraged the “one product, one process” character of today’s industry.
Early on, the need for advanced and integrated process expertise as a requirement for product development challenged the development of fabless MEMS companies. It also challenged foundries whose expertise and business model depended on a “many products, one process” approach to serve multiple customers necessary to amortize their capital investment. Accelerometers, microphones and optical components are extremely diverse products requiring unique process capabilities, application “know-how” and design tools. MEMs companies were forced to develop new products with their own custom process expertise and/or depend on a manufacturing partner for joint development of new products. Also, as most MEMs manufacturing was on 6” and less wafers with secondary equipment, there was no compelling reason to standardize key equipment types and their supporting integration requirements, as there are in leading-edge semiconductors.
Mike Rosa: Today, the bulk of all high volume MEMS are built on 200mm wafers in order to satisfy the volume/price requirement – this wafer size is currently under a lot of pressure due to ASP erosion 3-5% per quarter on MEMS devices. This has led a number of segment leaders to explore MEMS development on fully depreciated 300mm toolsets, not for the advanced line width capabilities of 300mm but simply because they can fit more devices per wafer. This combined with advanced packaging techniques will increase the number of die per wafer and reduce the per die footprint. This is currently considered the final incarnation of MEMS on wafer based technologies – beyond this, other manufacturing techniques such as roll-to-roll or large area substrate (akin to glass panel for flat screen TV) are being explored at the R&D level in an effort to support the latest drivers of mega-volume MEMS, such as wearable computing, trillion sensor vision and so on, which will require around two orders of magnitude price reduction over current MEMS devices with equivalent or increased device capability. With specific regard to the diversity in MEMS design/fabrication and its impact on their manufacturability and cost, the segment as a whole is now past this, to a large extent, with all participating vendors now owning the requisite manufacturing tools to make a variety of MEMS devices. Because of this the basic problem has reduced to number of die on a wafer and techniques used to make MEMS smaller (which means due to their requirement for minimum mass, either vertical integration of wafers/die or new sensing materials/techniques to replace current MEMS functions).
How are MEMS foundries standardizing MEMS fabrication processes to deliver faster turnaround, lower costs and reducing time to market for new MEMS device designs?
Mike Rosa: While there is, of course, a lot of discussion around standardization of MEMS processes, the reality is that very little can and is being done in this area. The foundries are usually followers in this space and work to adopt processes that can support the highest volume MEMS devices in the hope that they obtain work orders by second tier suppliers, fabless companies or overflow orders from IDMs. In recent times some of the ‘bright spots’ in standardization have occurred around the licensing of successful MEMS design architectures – fabless company Invensense is an example here, which licensed its fabrication process for Ge/Si integrated devices based on an advanced packaging technique.
Tom Morrow: As the MEMs market has grown, foundries have increasingly improved their processes, IP and technology necessary to support economical, high-volume production. Many MEM foundries have invested millions of dollars over the past several years in preparation for the emergence of a fabless MEMs market and are seeing the benefits of their efforts come to fruition. They have invested in process capabilities, PDK and design support, are moving from custom manufacturing to platform manufacturing flows, and have leveraged their relationships with packaging and test houses to compete with MEMs IDMs. Their standardization efforts are focused on internal or proprietary standards — how to move from custom flows to platform flows — to achieve economies of scale and work with their package and test partners on standard, repeatable, back-end business. MEMs process technology is also being increasing licensed further facilitating advanced MEMs capabilities by foundries.
Eric Mounier: We expect that as MEMS companies increasingly move beyond competing on manufacturing technology to competing on functionality, thus more of TSV / WLP packaging solutions will become widely-used platforms. A good example of an “in-house” front-end process standardization is the MIDIS platform from Teledyne Dalsa. Depending on the customer’s request, MIDIS has been developed to have a minimal change of design and process to answer the customer request. But Teledyne Dalsa is not the only one having such a standardized solution. Silex Microsystems and MEMScap have technology platforms. Other MEMS foundries with product platforms include Tronics and XFab.
Peter Himes: Today each foundry has their own efforts in this area. Some offer “standard platforms” but being able to use it depends on an exact fit of the intended design to the process flow. However, even these standard platforms should not imply they are standard across different fabs or foundries.
At Silex we have a methodology of standardizing at the block level which we call the SmartBlock approach. A process flow can be built out of two or three SmartBlock elements, plus product specific processing, to create the final process flow. This does not eliminate the process integration work or need to prove the process flow concept with actual silicon before starting any production qualification, but it does help substantially in de-risking the development, which translates in the long run to lower costs, faster time to market, and fewer unique process development efforts.