As the industry is pushing for further integration of CMOS and MEMS chips, a startup company Baolab claims that it has developed a process with which a variety of MEMS devices can be made on standard CMOS lines. We recently spoke with David Doyle, the company’s CEO, about the process’ advantages and limitations as well as competitive CMOS MEMS efforts from other companies and industry groups.
David Doyle: Our NanoEMS “inside CMOS” process is best suited for high volume, cost sensitive applications and for applications requiring a monolithic integration of the MEMS and the sensing electronic circuitry. Cell phone applications, such as RF tunable devices, are a clear market for this; here, monolithic integration to reduce space and join several functions inside the same IC, and above all cost, are of primarily importance.
The process flow includes (1) vHF to etch away the IMD through the pad openings, (2) plugging opening holes with additional deposition and (3) standard packaging.
MEMS Investor Journal: For which kinds of applications is your CMOS MEMS process not the best option? What are the fundamental limitations of your process?
David Doyle: Inevitably there are some limitations coming from the usage of a process, CMOS, which has not been conceived to make MEMS. CMOS can suffer from poor metallurgy, with no stress control and thin layers. The consequences of these limitations are curling up of metal layers, devices that are prone to creep, and reduced mass. In practice, this means that the MEMS devices will need to be small. This makes it a real challenge to design any useful MEMS devices. Baolab solves these challenges as they relate to our intended products by innovative designs, which are part of the company’s know-how and also the subject of many patents that the company owns. In several cases, an important part of the design effort is then shifted towards the electronics, which can be done because we have the monolithic integration. There are also other optional mask-less post-processing steps that can be added to improve the metallurgy for contact based designs. So, in the end, with this combination of design and process know-how, we are able to meet the target specs in most of the high volume, cost sensitive consumer applications.
Nevertheless, applications requiring a large mass or significant metal thickness would tend to be less suitable for our process, as would biocompatible applications such as some medical applications.
An
SEM image of a 1D accelerometer prototype that uses a novel algorithm
to detect accelerations down to the 10mG range, despite its low size
and reduced proof mass.
MEMS Investor Journal: Aside from fundamental limitations, what are some of the current process enhancements and challenges that you are working through?
David Doyle: It is important to stress we are working with standard CMOS processes and standard packaging, so “enhancements” that deviate from a standard CMOS process are not something we entertain. Modifying a CMOS process is something some others have done, but this directly impacts cost and ease of integration as additional process steps tend to be needed. The big challenge for Baolab in using standard unmodified CMOS was in working out how to make the etch process controllable, repeatable and predictable. Having solved that issue, the next challenge was in how to ensure standard package types can always be used. We strongly believe, and have proven over and over, that the lowest possible cost can only be achieved by using a standard unmodified process and standard packaging. With both of these main challenges largely resolved, the issues we are working through today mainly relate to yield improvement and design optimization.
An SEM image of a set of cantilever test structures used to characterize the stress of the different metal layers of the wafer, as they come from the standard CMOS process.
MEMS Investor Journal: What are some of the other companies that have developed or are developing CMOS MEMS processes? How does your approach compare to theirs? What are some of the pros and cons?
David Doyle: Intra-CMOS processes can really only be developed by companies which own their CMOS lines. As such, they need to re-qualify the process when implemented, and it cannot be easily transferred to other sites making it an expensive solution and limited in volume capability. It also needs to be reworked and re-qualified when upgrading to a lower node. Examples of this are well known -- the iMEMS process from ADI mainly used for inertial sensors, the holes made inside an IMD layer by Hitachi and Infineon, essentially to build pressure sensors, or the addition of tiny metal layers inside a microcavity by Cavendish Kinetics.
Post-CMOS processes like bulk etching of the substrate are limited to pressure sensors or similar types of devices. If they add metal layers on top of the CMOS wafer like WiSpry, they are adding several MEMS steps, making again the process more captive to a fab, pretty much like an intra-CMOS process, limiting volume capability and transferability, but most importantly increasing the cost, not only for the added steps but above all due to the packaging challenges for this type of solution. The post-processing approach developed at Carnegie Mellon University, used by Akustica and available also through the MPW MUMPS process from MEMSCAP (ASIMPS), that reuses the BEOL metallization like us, is more complex and involves several steps, thus substantially increasing the cost.
An SEM image of a MEMS switch.
MEMS Investor Journal: Which markets and applications are you targeting first?
David Doyle: Our primary focus initially is the mobile phone market and first applications will be motion sensors -- digital compass, accelerometers and 6-9 degree of freedom sensors. Engineering samples for these will be available towards the end of this year. Following these products we will be introducing a range of RF MEMS switches, these have been in development for a while now and engineering samples are expected mid 2011.
An SEM image of a metallic mesh.
MEMS Investor Journal: What are the specific objectives and milestones that you have outlined for the remainder of 2010?
David Doyle: Our main focus is delivery of engineering samples of motion sensors to potential customers in November. There are of course a number of key steps on the way to that as there are with any chip development but we are not ready to discuss those specific steps or the timing of those as we regard that as commercially sensitive.
*********************************************
David Doyle is a highly successful industry veteran with over 28 years experience, having started his career in semiconductor sales with the likes of Ferranti, Plessey and then Motorola. In 1991 he moved to European sales management roles in EDA with Intergraph and then VeriBest/Mentor Graphics. He joined the management team at ARC International in 1999 and helped in the company’s floatation in 2000, which was the biggest LSE IPO of that year. His last role at ARC was VP of Worldwide Field Operations until he was headhunted for the role of Baolab CEO in 2008 with the task of commercialising the company’s patented, disruptive NanoEMS technology for creating MEMS within a CMOS wafer.
Copyright 2010 MEMS Investor Journal
Comments