Expectations of the use of RF MEMS devices for mobile phone applications have gone through many cycles of hope and disappointment. We may now be finally entering in the mass adoption phase with the first high-volume application, WiSpry’s MEMS based tunable impedance matching device, incorporated in one of Samsung's new phones.
RF MEMS devices promise to bring significant benefits to both consumers and cell phone network operators. These benefits include fewer dropped calls, faster data transmission rates, improved reception and enhanced power efficiency. Use in mobile phones and tablets is projected to expand the RF MEMS market by a factor of 200 by 2015, according to IHS iSuppli. Sales of RF MEMS devices are expected to increase from $720,000 in 2011 to a projected $150 million in 2015.
DelfMEMS, an RF MEMS switch developer based in France, is expected to be among the players in this rush to commercialize RF MEMS devices for handset and tablet applications. MEMS Journal recently interviewed Olivier Millet, founder and CEO of DelfMEMS. We discussed the challenges involved in integrating RF MEMS devices into handsets, viable insertions points, product roadmaps, and competition from MEMS and non-MEMS technology providers.
MEMS Journal: Briefly, what is the history of DelfMEMS?
Olivier Millet: We founded DelfMEMS six years ago. We were involved in research on MEMS reliability and we were thinking that anchors were generating too much mechanical stress within the device structure. From this key aspect, we filed a patent for a flexible anchorless membrane.
Since 2006, we have been in continuous discussions across the entire value chain of mobile telephony, from the module makers to operators. This has allowed us to better understand the future challenges of this industry and to optimize our component to address these needs. We have been able to deliver custom samples to a leading telecom operator last year.
MEMS Journal: RF MEMS has not found an application in handsets for the past decade. What is different now?
Millet: The difference lies in two points. The first is RF MEMS technology itself. The technology is better understood, with significant improvements in terms of switching time, thin film packaging for module integration, higher lifetime reliability with the possibility of the hot switching operations (e.g. in billions of cycles). MEMS foundries can now also provide, within the required high capacity, process technologies for optimal performance in terms of insertion loss, linearity with a low cost.
The second point is the context. The arrival of 4G with protocols such as TD LTE (Time Division, Long-Term Evolution, a 4G mobile telephone standard) is a strategic turning point for mobile players. The players need to position themselves to have access to a technology to control or even reduce the level of losses, to deal with an increasingly limited spectrum, to integrate future modules and to manage the bill of materials, as well as to continually bring power consumption downward.
MEMS Journal: What are some of the main design challenges for RF MEMS in mobile handsets? How are DelfMEMS and others handling those challenges?
Millet: There are two types of RF MEMS devices -- digital capacitors and ohmic switches. DelfMEMS is currently focused on the ohmic switch and our design of experiment (DOE) are currently running for digital caps within the same process.
The ohmic MEMS switch is a competitor of SOI, SOS and pHEMT technologies. It offers ultralow loss and stellar linearity. The main design elements of a MEMS switch are the contact force, the restoring force and the contact material. The flexible anchorless membranes have the ability to generate large contact forces for low losses, and significant restoring forces to perform hot switching. This type of design also allows us to have a fast switching time less than a microsecond with a controlled and limited bouncing effect.
Anchorless design limits the temperature dependence, vibration, shock and lower sensitivity to stress induced by the packaging, and drastically improves the yield (less material and design constraints). The packaging (e.g. switch capping) itself must be compatible with overmolding for final integration. Finally, the overall die size has been minimized to address the low cost requirement.
MEMS Journal: RF MEMS have so far not penetrated the consumer electronics market due to cost and high-volume production issues. How have those issues been addressed?
Millet: MEMS technology itself is not necessarily an expensive technology. The minimum step for lithography is about 5µm and without using an expensive DRIE process on high SOI wafers compared to typical inertial sensors. In fact, cost drivers are quite similar to IPD technology.
Another important point is the achievable yield. RF MEMS switches are quite different from current MEMS products such as sensors. They are in fact surface micromachined actuators and their functionality is primarily linked to the quality of the contact resistance, less dependent on the dimensional precision or the uniformity of the material properties and thickness. It was a similar situation for devices such as DLP, with millions of devices per die but with stellar yield.
DelfMEMS has already implemented the RF switch process flow on top of existing processes (such as IPD), but we will provide at the beginning a two-die solution with the ASIC (charge pump and driver) on a separate die to insure a more manageable and cost effective integration. DelfMEMS’ roadmap is to propose post-process of the IC in 3-4 years when the yield will be a CMOS-like.
DelfMEMS owns the manufacturing process and has a dedicated team to continuously improve it. We think this is a crucial point to manage well in order to achieve optimum manufacturability. For volume production, we already have partnered with a foundry with a capacity of hundred millions of units, and we will be able to transfer to a second source for multisourcing.
We are fully aware that the current trend is to go to a CMOS monolithic process. First, we don’t see full MEMS CMOS process compatibility; a better approach is a modified CMOS process developed around a specific line of the foundry. From there, integration of a contact material fitting with lifetime requirements of the device is a key challenge.
MEMS Journal: How do RF MEMS address the issue of multiple standards in smart phones and in antenna tuning problems that result in dropped calls?
Millet: DelfMEMS is in discussions across the complete mobile value chain in order to better understand the constraints generated by high data rates and a crowded spectrum. The architectures that RF module and chipset makers are considering as prime for MEMS switch applications involve schemes for duplexer elimination/simplification, converged-mode amplifiers and 4G designs for some of the more critical LTE mode specifications that lie ahead of them.
They believe that a difference of 0.3dB in insertion loss is an extremely important in future 4G systems, where linear power will be at a premium. Although this saving in linear power seems small, it implies a significant change in architectures. There are more and more switch paths between active devices such as power amplifiers and the antenna. Insertion loss of switches in series is accumulated and accelerates degradations of performances of the overall system.
The decrease of intermodulation and harmonic distortion is also a key point. The power in full duplex has to be delivered through a highly specified (linear) RX/TX switch. The half-duplex switch needs low harmonics. The design issues of the PA and switch paths are well understood, but if both full and half duplex paths need to be supported then the design and implementation becomes problematic, particularly when all the other existing switch paths are taken into consideration. Filter banks, tunable filters and tunable PA are also addressed by combining arrays of MEMS switches with IPD.
MEMS Journal: What makes your RF MEMS switches and relays unique?
Millet: Our RF MEMS technology based on an innovative mechanical design: the flexible anchorless beam. Our unique patents portfolio allows addressing past issues in term of high contact and restoring forces, and stability over integration, temperature, shock and vibration resistance. This new type of design enables downscaling of our components to fit with size and cost requirements.
Moreover, our switches are broadband from DC to 7 gigahertz, and therefore are a match for mobile communications. We are finalizing the development of these unique building blocks, which can be easily used and duplicated to provide products without a large increase in parasitics.
This building block switch will also fit with the antenna, mode and tuning requirements. Our RF MEMS switches are like an added layer on a low cost MMIC design, with low NRE for development of new product.
MEMS Journal: Can you tell us about your competition? What are the other MEMS switch companies in the market and how does DelfMEMS compare with them?
Millet: For mobile, current competition could be seen as pHEMT, SOI technology and SOS technology. Actually, MEMS are not in competition with these technologies as the added value relies in the combination of switching technologies. MEMS switches offer ultralow loss, increasing the gap with other technologies when you increase the number of throws or the number of switches in series. MEMS switch also offer an incredible linearity. It's a dedicated technology for LTE applications, enabling the use of several channels simultaneously.
Mobile is a driver for DelfMEMS technology. We propose a building block used for a lot of applications with low cost, hot switching, high integration and high lifetime. The point is that MEMS is broadband and this building block can be used for ATE, instrumentation, automotive radars at 24GHz, as well as military and space applications. Using our packaging technology, we can limit the range of frequencies, up to 35GHz.
MEMS Journal: What about other RF MEMS switch makers?
Millet: Based on what we know, other galvanic MEMS switching companies are XCOM Wireless (U.S.), Radant MEMS (U.S.), and OMRON (Japan). OMRON is working on a new version of its MEMS switches for mobile applications. We strongly believe that we have the best solution for ohmic switches for mobile applications in terms of hot switching, switching time, integration, process windows and cost.
MEMS Journal: What are the most likely insertion points for MEMS switches in cell phones and base stations? What is the business driver for each? Which insertion point is DelfMEMS targeting first and why?
Millet: DelfMEMS is led by customers. We have many inputs from different markets, but less in the base station area. It may be that the need for MEMS switches is currently not strategic for these applications. For mobile, one potential insertion point is the antenna switch. It does not increase the bill of material of the front end as we are replacing an existing device, and it should even drastically simplify the radio-frequency architecture (number of duplexers). That's also the first target of DelfMEMS for this market. DelfMEMS will deliver multi-throw MEMS devices to be integrated within front-end modules. The reason is that we have the customers for this application and our technology fits with their expected specifications.
MEMS Journal: What are your future plans and expected revenue and market share?
Millet: DelfMEMS is one of the leaders in MEMS switches for telecommunications. Within the supply chain, DelfMEMS delivers bare die to module makers, which are then integrated across a range of modules such as the ASM (antenna switch module).
DelfMEMS has a step-by-step approach. Our first product will be a combination of switches for RF front-end for mobile applications extending toward digital caps and ohmic switches within the same process. DelfMEMS is also working on using the same building blocks for other markets with the same broadband technology (up to 30GHz) without new development costs except for the integration into the QFN packages.
Our goal is to establish a strategic market position with the most compelling RF switching product platform on the $1.5 billion low-power switch market.
This article is a part of MEMS Journal's ongoing market research project in the area of RF MEMS switches. If you would like to receive our comprehensive market research report on this topic, please contact Dr. Mike Pinelis at firstname.lastname@example.org for more information about rates and report contents.
Copyright 2012 MEMS Journal, Inc.