Home


Feature article Think Small! issue 4, Volume 1:

Has the time come for MEMS oscillators?

by Jérémie Bouchaud, Bernardo Knoblich

A flurry of announcements during the year have resurrected interest in rf mems oscillators. Wtc believes that these devices have a nice future, but warns of some confusion and inflated expectations.
MEMS oscillators are nothing new. The first publications mentioning rf mems resonators for oscillator applications appeared in the 1980s. Until recently however, the low temperature stability of silicon and the need for expensive ceramic or metal vacuum packaging prevented mems from becoming a serious alternative to quartz.
In the last five years, tremendous developments in processes, packaging and integration of circuitry have taken place, led mainly by startup companies like Discera, Silicon Clocks and SiTime. Discera and SiTime have samples available now. SiTime plans serial production for the last quarter of 2006.
Before we go any further, let's clearly describe a mems oscillator. We have noticed some confusion between oscillators and resonators, especially when discussing price. A mems oscillator chip consists typically of 1) the resonator, 2) the circuits for oscillator function, frequency compensation and temperature compensation, and finally 3) the package. The micromechanical resonator is a micromachined device that vibrates at a specific frequency due to an external excitation (mainly electrostatic to ensure a high quality factor).
The market figures given in this article (e.g. chart below) concern mems oscillators and not only mems resonators, since WTC expects most suppliers to provide a full oscillator.

The resonator is inexpensive in volume and is estimated to be less than $0.10. The typical price of a mems oscillator will start at $0.50 (see diagram below), and may reach a few dollars in some instances.

Oscillators in today's market Early suppliers of MEMS oscillators claim they can address 50%-90% of the clock and oscillators market. This is a mass market estimated in the range of 10 billion units and amounting to $3.2 billion. We differentiate mainly between three types of products:

  1. The clock oscillator: A circuit that creates a series of pulses that pace the electronic system. According to the different manners of implementation, clock oscillators can be divided into xo (crystal oscillators), vcxo (voltage controlled crystal oscillators), tcxo (temperature compensated), ocxo (oven controlled), and vcso (voltage controlled saw oscillators)
  2. The clock generator: A clock oscillator with more complex arrangements. The basic parts that are shared by clock generators are a resonant circuit and amplifier, and they may have additional sections to modify the basic signal. Clock generators in this article refer to oscillators with multi-plls (Phase Lock Loop) or jitter cleaners.
  3. The real time clock: A clock that keeps track of the time even when the system is turned off.

Currently, the main technologies for clock products are quartz, ceramic and cmos silicon clocks. The market is mostly served by quartz (70%-80% of the value), and Japanese companies such as Kyocera, ndk and Toyocom dominate. Quartz technology has been around for decades, exhibits very good temperature and long term stability and is therefore ideal for high precision oscillators. The frequency accuracy, including all the environmental effects and one-year aging, is of the order of few ppm to tens of ppm.
Silicon (i.e. non-mems, simply electronic) timing or cmos clocks are available from Maxim or Linear Technologies and gradually replace ceramic devices in applications with relaxed specifications. Worth noticing are also new kinds of inductive (lc) cmos clocks such as those from Mobius Microsystems. These devices target new architectures with a frequency reference in the gigahertz range instead of the megahertz range. Their frequency accuracy is now better than 500 ppm.
Mems oscillators fit between cmos clocks and quartz oscillators.

MEMS: theory and practice
The advantages of mems are well known and found in numerous articles over the past year. Therefore we briefly summarize them in order to expose certain inconsistencies in the arguments.
Mems technology offers in theory major advantages in terms of size (especially for mobile applications), power consumption, cost and ability to provide multiple operation frequencies. In practice however, it appears that the first generation of rf mems oscillators will not initially fully capitalize on the mems' smaller size or improved capacity for integration. So why and when to use mems?
Size. A tiny mems oscillator is often pictured on top of a 100 times bigger quartz oscillators. However, one should also acknowledge the latest shrinkage in some quartz products. A new generation of quartz tcxo is emerging with similar dimensions to the first mems oscillators. Indeed the first mems oscillators will be pin-to-pin compatible with quartz products to allow one-on-one replacement. Since the size of the package will be the same, the main driver for mems oscillators in the "small oscillator segment" is price since manufacturing and encapsulating of small quartz devices is expensive compared to mems. Pcb space can also be saved with mems oscillators, which do not require additional bypass capacitors and resistors.
Low cost from batch manufacturing. Mems is cheaper than quartz but we think it will not compete in mainstream markets, rather those where they obtain a premium for the size, higher frequencies and power consumption.
CMOS integration. The first mems oscillators will actually be hybrid parts, e.g. SiTime's mems resonators are wire bonded to the asic, as monolithic integration of the mems resonator with the ic does not make economic sense when the mems die is much smaller than the ic.
Low power consumption. This advantage is also diminished in a hybrid part due to wire-bond parasitics.

Applications roadmap
Early on, the product with the most potential was not always clear. A tcxo meeting tough requirements in temperature stability and phase noise was first announced around 2002, but did not materialize.
In 2004, people said real time clocks used in stand-by functions at 32 kHz would be first to market; this also did not emerge. Now, a consensus has formed around the following roadmap:

Taking a longer term perspective, there is a strong r&d effort on integration and new architectures that use resonators as filters or mixers. Among these, stmicroelectronics' vision of embedded nems leveraging new manufacturing processes such as the Silicon- On-Nothing (son) is interesting. Clark Nguyen, the mems resonator pioneer at the university of Michigan, is also pushing the development of the "all mems receiver front-end" resonator, allowing for full tunability from 600-2500 MHz. Also worthy of note are developments in carbon nanotube resonators, currently explored at epfl in Switzerland.

Markets and key players
So how will it all shape up? Overall, we expect the market for mems oscillators to in the range of $200 million in 2012. It is far from the 50% to 90% of the market that can be addressed in theory, but it is still a very interesting opportunity which justifies all the excitement. The key will be again to target the specific applications where mems oscillators can obtain a premium, for example where small packages, low power consumption, or higher frequencies are required, or in harsh environment applications and eventually for higher integration with ics.
Among key players are three us start-ups: Discera, SiTime, and Silicon Clocks. The first commercial serial products are expected from SiTime and Discera. These companies are fabless and will sell devices since this model allows for higher revenues than licensing. Large semiconductor companies like nxp, st and potentially Freescale are also likely to join the fray.
The mems sensor company vti Technologies recently announced its cooperation with the leading Finnish research organization vtt and mems products are anticipated in 2008. Large Japanese manufacturers of quartz oscillators such as Epson are investigating mems oscillators, and some may enter the market by 2010. Some quartz vendors are currently also considering acquisition of mems technology.


A note of caution With so many nice things going on in mems oscillators, we offer two notes of caution: first, specifications and second, reliability. Regarding specifications, one needs to make sure the mems oscillator works at sufficiently high temperatures. At the same time, reliability testing requires effort and time spent on the production of devices rather than of research wafers.

Conclusion: hype or hit?
During the 2002-2004 time frame, we polled potential end-users in the cell phone and consumer electronics industries for their views on mems. Most did not consider it necessary to abandon established quartz technology. Today, the same people talk with enthusiasm about mems oscillators. This goes to prove that several years of developmental engineering have paid off handsomely and solved many of the early commercialization growth pains. New, exciting technology targeting mass markets driven by start-ups are often characterized by confusion or exaggeration, but all in all, we believe that mems oscillators will follow bulk acoustic wave devices as the second rf mems mass product. If SiTime keeps to schedule and ships its first million devices by years-end, a major step will have been taken towards a bright mems future!